From the frontier AI front,

• The Wall Street Journal reports,
  • “Nvidia NVDA is planning to use a $6 billion deal it struck this week [August 22] to build one of the world’s most powerful open-weight AI models, one that would compete with Chinese heavyweights like DeepSeek and Kimi K3, according to people familiar with the matter.
  • “The chip giant’s licensing deal with the AI startup Poolside is also set to present a direct challenge to frontier U.S. AI companies including OpenAI and Anthropic, since open-weight models are generally far cheaper to operate and allow easy customization.
  • “Although Nvidia counts big labs like OpenAI and Anthropic as some of its closest partners, the move reflects Chief Executive Jensen Huang’s efforts to position Nvidia for success in multiple AI battlegrounds simultaneously, ensuring the company is able to hold on to its dominant market position.”
• and
  • Nvidia and CrowdStrike are addressing what they see as one of cybersecurity’s biggest gaps: the need for cyber defenders to harness AI at the velocity of cyber attackers.
  • In the current cyber landscape, frontier AI models are discovering bugs at machine speed—raising alarm bells among experts, who warn that when autonomous hacking models go rogue or attack companies, chaos is on the horizon.
  • But cyber defenders aren’t getting enough of that same AI firepower, according to Nvidia and CrowdStrike, who on Tuesday [September 1] introduced a new family of agentic AI models, dubbed SafeMind, which can both find attack paths and close them for customers. The models are available in CrowdStrike’s flagship Falcon platform and through its API. * * *
  • “SafeMind has two models: Red Tempest, which is an “offensive” model that emulates AI adversaries, and Blue Solano, which is a “defensive” model designed to fix identified issues. The two models are connected by software called a harness, which runs the models in a closed-loop system that pits them against each other for self-improvement.”

• The New York Times reports,
  • “Nvidia said on Thursday [September 3] that it is buying Hugging Face, a library of open artificial intelligence models, for $12.9 billion, as the chipmaker extends a spending spree in the escalating global race to dominate the technology.
  • “The deal marries Nvidia’s chip and data center infrastructure with Hugging Face’s millions of open-source A.I. models that can be freely downloaded and modified. And it puts Nvidia’s deep pockets firmly on one side of a debate over how A.I. systems should be built.
  • “Together, we will make A.I. more open, more capable and more accessible to people and institutions around the world,” wrote Jensen Huang, the chief executive of Nvidia, in a blog post.
  • “The acquisition is also another sign of Nvidia’s growing role as Silicon Valley’s central banker, using its vast financial resources to bolster A.I. start-ups and funnel money to customers for its chips. With $197 billion in current assets and nearly $60 billion in profits from its most recent quarter, the Silicon Valley giant has been spending heavily on A.I.”

• Daniel Borish writing in LinkedIn lets us know,
  • “Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 this week, positioning them as its most capable models yet for coding and knowledge work. The two are the same underlying model, split only by the safeguards wrapped around them. Fable 5.1 is generally available. Mythos 5.1 is restricted to vetted cybersecurity and life sciences professionals through trusted access programs run in partnership with the US government.” * * *
  • “On the cybersecurity side, Claude Code users should see roughly 60% fewer safeguard interventions per session, and Fable 5.1 can now be used to find software vulnerabilities, though not to write exploits for them. Tasks like penetration testing, exploit generation, and binary-based vulnerability scanning still route to Anthropic’s Opus models. On biology, updated safeguards reportedly fire 85% less often on benign medical and elementary biology questions, though research-grade life sciences queries still get redirected to Opus, and full access to Mythos 5.1’s biology capabilities is limited to the government-linked Life Sciences Verification Program.
  • “Anthropic’s safety testing found Mythos 5.1 better aligned than its predecessor on several measures it tracks: less likely to seek resources outside its assigned environment when given an impossible task, less likely to reason its way around explicit constraints, and less likely to attempt or succeed at reward hacking. The company also disclosed limits to that picture, noting its behavioral audits have less visibility into very long-context work, multi-agent settings, and testing found the model can still sometimes bypass approval steps and automated review classifiers.
  • “Fable 5.1 also ships with new anti-distillation measures. New API accounts created from launch onward can no longer edit Claude’s prior context in a conversation while preserving the model’s recorded thinking, closing off a publicly documented technique used to extract that thinking at scale.”
    
• The Wall Street Journal adds,
  • Anthropic has signed a cloud-computing deal worth $35 billion with NvidiaNVDA 3.21%increase; up pointing triangle-backed cloud provider Lambda, with Nvidia itself holding the lease on the data center, according to people familiar with the deal. 
  • The data center is being developed by Hut 8, a bitcoin miner and data-center developer, in Nueces County, Texas. Nvidia signed an agreement with Hut 8 a few weeks ago to secure the capacity, the people said. 
  • The convoluted arrangement is another example of Nvidia’s growing role in helping non-investment-grade firms such as Anthropic get access to its expensive computing resources. The deal also helped a nascent cloud provider, Lambda, secure a lucrative contract with Anthropic without needing to procure the data-center space on its own. 
  • Lambda will use the data center Hut 8 is developing to plug in chips bought from Nvidia, which is also its investor. It is not clear how much Lambda will pay Nvidia to access the data-center space.
  • Anthropic has been racing to sign cloud-capacity deals after hitting a supply crunch earlier this year that coincided with its products gaining traction. The AI developer signed a $45 billion deal earlier this month with another Nvidia-backed neocloud, Nscale, to rent Nvidia capacity from its West Virginia site, according to people familiar with the deal. 

• Security Week relates,
  • OpenAI said its newest model, Astra, has reached the ‘Critical’ cybersecurity capability level under the company’s Preparedness Framework, the first time any of its models has been placed in that category. 
  • The designation applies when a model can independently find and exploit zero-day vulnerabilities across many well-defended systems, or carry out a complete cyberattack against a hardened target from only a high-level instruction. OpenAI said the classification requires additional safeguards before the model can be released.
  • In testing described by the company, Astra achieved a perfect score on ExploitBench, a benchmark that measures a model’s ability to turn known vulnerabilities into working exploits. During a separate evaluation involving more recently disclosed flaws, Astra uncovered two zero-day vulnerabilities on its own. 
  • The model also broke out of a browser sandbox to run commands on the underlying machine, and separately chained several flaws in a hardened operating system to gain root-level access.
  • OpenAI reported that Astra now declines 91.5% of cyber-related jailbreak attempts in its testing, up from 59% for its predecessor, GPT-5.6 Sol. The company also said Astra showed far less tendency than Sol to bypass safety restrictions or take advantage of deliberately placed “honeypot” targets during evaluations. 
  • Full cybersecurity capabilities will not be widely available at launch. OpenAI plans to give a group of testers early access, with wider availability to follow through its Daybreak Blue program.

• Cybersecurity Dive adds,
  • “OpenAI on Thursday [September 3] said it would commit $1 billion in subsidized access and training to help small IT security teams use frontier AI to protect essential services, including electricity, water and local government services. 
  • “The Daybreak for Frontline Defenders program will be used to help defenders search for critical vulnerabilities in their software, hunt for suspicious activity in their technology stacks and stop malicious actions if hackers are able to gain access to their systems.
  • “OpenAI plans to roll out a six-month pilot program with the Multi-State Information Sharing and Analysis Center (MS-ISAC), which will train and support a group of water utilities and other public sector defenders.” 

• Dark Reading points out,
  • “A new study suggests that while AI is rapidly accelerating vulnerability discovery, enterprise organizations are better equipped to handle the surge than generally assumed.
  • “The key is their ability to quickly validate findings, prioritize risk, and get available fixes into production.
  • “Software supply chain security firm Echo recently analyzed nearly 40,000 CVE life cycles across 250 open source container projects, drawing on a year of its own platform telemetry, survey responses from more than 80 security leaders, and an independent analysis of Anthropic’s Claude Mythos.
  • “The results, detailed in a report titled “Mythos Readiness Report,” show how AI is transforming vulnerability discovery and exploit development — something that security teams have been encountering firsthand over the past year.”
• and
  • “With multiple benchmarks confirming that at least one frontier model can autonomously execute an end-to-end compromise, cybersecurity experts are warning that companies have to pick up the pace in securing their attack surfaces and adopting AI-speed defenses.
  • “On Sept. 2, consulting firm Booz Allen became the latest organization to confirm that a frontier model — Anthropic’s Mythos 5 — can act as a fully autonomous hacker and compromise a production-grade enterprise network. The company also released a metric, dubbed the Cyber Weapon Index (CWI), to benchmark a model’s capabilities, pairing the ability to find and exploit vulnerabilities with the ability to execute and attack a target.
  • “In their evaluation, Mythos scored an 80, while the next-closest contender, SpaceXAI‘s Grok-4.5, scored a 49. However, the current standings are not as important as where we will be in six months, says Brad Medairy, president of Booz Allen’s National Cyber practice.
  • “Our view is there’s going to be some level of parity, at least between the frontier models and the Chinese models over probably a six-month horizon, and so everyone is in a rapid race to evolve these capabilities,” he says, adding that, when facing human attackers, defenders had the advantage, but “in the future, the balance of power shifts to the offense, because the attacker can deliver effects at speed and scale that the traditional defenses can’t keep pace with.”

From the cybersecurity policy and law enforcement front,

• Federal News Network reports,
  • “Congress is kicking the can down the road again on a long term debate over reauthorizing the Cybersecurity Information Sharing Act of 2015.
  • “The continuing resolution passed by the House on Tuesday would extend CISA 2015 through the stopgap funding period into early December. The Senate has already passed the measure, meaning it now heads to President Donald Trump’s desk. [FEHBlog note: The President signed the continuing resolution into law on September 2.]
  • “There has been a persistent chorus of voices in industry calling on Congress to find a long-term solution to re-authorizing CISA 2015. Those calls have only grown louder amid advancements in artificial intelligence models and recent cyber attacks on critical infrastructure.
  • “The 2015 law provides privacy and liability protections to encourage companies to share data about cyber vulnerabilities and threats with government and each other. The information sharing law briefly lapsed during last fall’s shutdown and again earlier this year before being extended through Sept. 30.”

• The Wall Street Journal points out,
  • “The U.S. Defense Department is expected to release its cyber plan as early as Tuesday, NextGov reported, citing people familiar with the matter. The plan is due to outline how the Pentagon will integrate cyber tactics into military strategy, emphasizing offensive actions, and address training requirements. The most recent Defense Department cyber strategy was issued in 2023.”

• Per a CISA news release,
  • “CISA and the Group of Seven (G7) Cyber Security Working Group released Preparing for the Post-Quantum Era: A Call to Action highlighting the urgent need for organizations and governments to begin transitioning to post-quantum cryptography (PQC) to protect sensitive data, authentication systems, and critical assets from emerging quantum computing threats.  
  • “The G7 Cyber Security Working Group’s call to action outlines five priorities for a successful transition to PQC:
    • “Raising awareness of quantum risks and the importance of PQC;
    • “Developing national strategies that support PQC adoption and integration;
    • “Advancing research and development for quantum-safe technologies;
    • “Fostering public-private partnerships to share expertise and resources; and
    • “Integrating PQC into cybersecurity requirements and procurement processes.” 

• Cyberscoop relates,
  • “The Trump administration rolled out a six-month test program in Texas on Monday that will draw on volunteer expertise and technology from cyber and artificial intelligence companies to protect the water sector.
  • “Project Watershed 250” is the first of the state-based, industry-centric pilot projects to cross the finish line that the Office of the National Cyber Director has previewedsince the publication of its national cybersecurity strategy earlier this year. Water systems have long been viewed as among the most vulnerable and neediest critical infrastructure sectors, and in recent months the sector has been the victim of a spree of attacks.
  • “Project Watershed 250 is a commitment from the states, industry and federal government that we will continue to prioritize our nation’s safety and deliver on America-first policies for the American people,” National Cyber Director Sean Cairncross said at a rollout event in San Antonio Monday.”
• and
  • “Sen. Ron Wyden, D-Ore., is asking the National Security Agency to update public guidance on the security risks associated with commercial virtual private networks, and to answer questions about foreign surveillance threats against standard VPNs.
  • “In a letter to NSA Director Gen. Joshua Rudd that Wyden sent Wednesday, the senator continued his push to warn about standard, commercial VPNs, following letters to federal agency leaders in March and July.
  • “While commercial Virtual Private Networks (VPNs) are recommended by federal agencies and widely marketed as shields against online spying, standard consumer VPNs do not sufficiently protect users from sophisticated adversaries,” Wyden wrote in the letter, first reported by CyberScoop. “Other, more secure alternatives are widely available.”
  • “Wyden took aim at “single-hop” VPNs that routes data through one server before arriving at the destination.”

• Cybersecurity Dive tells us,
  • “The Cybersecurity and Infrastructure Security Agency (CISA) is scaling back the free assessments it offers to critical infrastructure organizations, a move that marks a significant retreat from the agency’s core mission of helping secure the nation’s infrastructure.
  • “CISA’s regional staff will no longer perform its Cyber Resilience Reviews, Cyber Resilience Essentials surveys, Ransomware Readiness Assessments, Incident Management Reviews, External Dependencies Management Assessments or Cyber Infrastructure Surveys, the agency confirmed to Cybersecurity Dive.”
• and
  • “U.S. law-enforcement agencies and the cybersecurity firm CrowdStrike took down a 23-year-old Russia-based botnet on Monday.
  • “Authorities from the Justice Department, the FBI and the Defense Criminal Investigative Service seized U.S.-based domain namesbelonging to the Sality botnet, while CrowdStrike analysts worked with the agencies to sever infected computers from the botnet. Investigators in Bulgaria, Hungary and Romania also took down Sality domain names in their jurisdictions.
  • “Since 2003, Sality powered spam campaigns, credential-theft operations, distributed denial-of-service (DDoS) attacks and malicious proxy networks. In 2018, the botnet began deploying malware that hijacked cryptocurrency transactions by replacing copied wallet addresses with attacker-controlled ones.
  • “Sality “has been one of the most persistent threats on the internet, not because of its payloads but because of the robustness of its architecture,” CrowdStrike said in its report on the botnet’s takedown.”

From the cybersecurity breaches and vulnerabilities front,

• Dark Reading reports,
  • “Social engineers are trying to convince companies to make large-dollar transfers to their own accounts, under the guise of fake merger & acquisition (M&A) deals.
  • ‘It’s one of the oldest gambits in cybersecurity — the advance fee scam — for a new generation, and with more on the line. Gen, the parent company of cybersecurity brands Norton and Avast, was targeted by a ruse masquerading as a corporate acquisition. Gen investigated the attempted attack and discovered the company was just one of at least five targets, all of which were at risk of losing massive sums of money.
  • “Social engineers are trying to convince companies to make large-dollar transfers to their own accounts, under the guise of fake merger & acquisition(M&A) deals.
  • “It’s one of the oldest gambits in cybersecurity — the advance fee scam — for a new generation, and with more on the line. Gen, the parent company of cybersecurity brands Norton and Avast, was targeted by a ruse masquerading as a corporate acquisition. Gen investigated the attempted attack and discovered the company was just one of at least five targets, all of which were at risk of losing massive sums of money.”

• HelpNetSecurity relates,
  • “Attackers are targeting prominent individuals, their relatives and personal contacts to gain persistent access to their accounts, including private emails and files, the FBI has warned.
  • “The FBI’s Internet Crime Complaint Center (IC3) says the activity, which uses a technique called “OAuth consent phishing,” has been ongoing since late 2025. The FBI describes it as “a deceptive, sophisticated approach to access user accounts without requiring a password.”
  • OAuth is a framework that lets one application request access to a user’s account on another service without the user handing over login credentials directly.
  • “Recently observed activity includes impersonating government officials, media, and other publicly known personalities on a commercial messaging application (CMA) and soliciting the targeted individual to access a malicious link under the guise of a file sharing service through an application under the malicious actor’s control,” the FBI wrote in its advisory.”

• Dark Reading tells us,
  • “A coordinated operation has targeted at least 150 Microsoft Teams users across multiple companies with voice phishing (vishing) attacks aimed at installing remote monitoring and management (RMM) and malware tools onto their machines. The campaign also attempts in some instances to compromise organizations’ domain controllers.
  • “Researchers from Palo Alto Networks observed the operations — which they’ve dubbed “Spring Ring” — between January and April attacking employees across at least 10 organizations, according to a report published this week. The campaign illustrates a growing shift away from traditional email phishing toward attacks conducted through trusted enterprise collaboration platforms, which adds authenticity to the interaction.
  • “The Spring Ring operation represents an evolution from previous campaigns by merging vishing into the Teams workflow,” Noam Sala, a Palo Alto Networks staff researcher, wrote in the post. “This shift moves the attack from a passive click-and-harvest model to a real-time engagement.”

• The Cybersecurity and Infrastructure Security Agency added ten known exploited vulnerabilities to its catalog this week.
  • August 31, 2026
    • CVE-2026-81578 PaperCut NG/MF Missing Authentication for Critical Function Vulnerability 
    • CVE-2026-82078 PaperCut NG/MF Unsafe Reflection Vulnerability
      • The Hacker News discusses these KVEs here.
  • September 2, 2026
    • CVE-2026-9586 Sangoma Switchbox SQL Injection Vulnerability
    • CVE-2026-48710 Kludex Starlette HTTP Request/Response Smuggling Vulnerability 
    • CVE-2026-49869 Kestra OSS OS Command Injection Vulnerability 
    • CVE-2026-59822 BerriAI LiteLLM Improper Authentication Vulnerability 
    • CVE-2026-82329 JFrog Artifactory Improper Authentication Vulnerability 
    • CVE-2026-83548 SonicWall SMA1000 Appliances Server-Side Request Forgery Vulnerability 
    • CVE-2026-83549 SonicWall SMA1000 Appliances OS Command Injection Vulnerability 
      • Cybersecurity Dive discusses the SonicWall KVEs here.
      • ePlanetSecurity discusses the Sangoma KVE here.
      • AssurePort discusses the Kludex KVE here.
      • Sentinel One discusses the Kestra KVE here.
      • CSurface discusses the BerriAI KVE here.
      • Security Week discusses the JFrog KVE here.
    • September 4, 2026
      • CVE-2026-85046. Google Chromium V8 Type Confusion Vulnerability\
        • The Hacker News discusses the Google KVE here.

• Security Week tells us,
  • “Anthropic has warned some Claude users that infostealer malware on their computers allowed attackers to hijack login sessions and run up usage limits, according to an email the company sent to affected customers. 
  • “The company said it detected the activity, signed out the compromised sessions, and removed saved payment methods from affected accounts as a precaution.
  • “Anthropic is alerting Claude AI users whose computers were infected with infostealer malware such as Vidar, Lumma, StealC, RedLine, and Acreed on Windows, along with Atomic Stealer (AMOS) on a small number of macOS devices.
  • “The AI giant emphasized that the malware is general-purpose and not tied to Claude itself, typically arriving via unofficial downloads or malicious apps.” 

From the ransomware front,

• HIPAA Journal reports,
  • “Data breaches have been confirmed by Alta Orthopaedics in California, Cornerstone Behavioral Healthcare in Maine, Cameron Regional Medical Center in Missouri, Suntree Internal Medicine in Florida, and Associated Endocrinologists in Michigan. Ransomware groups have claimed responsibility for the attacks.”

• The Register adds,
  • “US hospital operator Nutex Health says attackers stole private or confidential patient, employee, provider, business, and financial information during the cyberattack it disclosed last week.
  • “In an updated filing submitted to the Securities and Exchange Commission (SEC) on Monday, Nutex also said an unauthorized third party had threatened to publish the stolen information.” * * *
  • “Nutex did not identify the intruders. However, The Gentlemen ransomware-as-a-service (RaaS) operation added the company to its leak site on Monday and claimed responsibility for the attack.
  • “The gang offered no evidence or details to substantiate its claim. Naming victims on a leak site and threatening to publish their data is a standard pressure tactic in double-extortion attacks.”

• InfoSecurity Magazine relates,
  • “Aurora ransomware actors have been observed abusing SpaceX’s AI Cursor Agent as part of exploitation campaigns, according to a new study by Gambit Security’s Threat Intelligence team.
  • “The threat actors ran Claude Sonnet through Cursor Agent to assist with various exploitation activities against 10 victims between April 8 and May 26, 2026.
  • “These tasks included scanning the victim’s environment for reconnaissance purposes, installing a VPN client and running certificate attacks.
  • “While Cursor Agent did not always achieve its stated objectives, the research demonstrated how threat actors are continuously experimenting with AI tools to speed up and enhance their campaigns.
  • “Cursor Agent is used by software developers to complete complex coding tasks independently, run terminal commands and edit code.”

From the cybersecurity business and defenses front,

• The Wall Street Journal reports,
  • “Palo Alto PANW Networks swung to a loss in the fiscal fourth quarter, but posted higher revenue and projected double-digit revenue growth in the current fiscal year as customers continue to boost their cybersecurity spending in response to advancements in artificial-intelligence.
  • “The cybersecurity company on Tuesday [September 1] said it is expecting revenue to rise between 23% and 24% to $14.1 billion to $14.2 billion in fiscal 2027. Analysts polled by FactSet are expecting $13.84 billion in revenue.
  • “Chief Executive Officer Nikesh Arora said the results and outlook reflect growing recognition among enterprises that they need to modernize their cyber defenses to meet the capabilities of ever-more-powerful AI models, especially following the release of Anthropic’s Mythos earlier this year.
  • “In that context, people are gravitating towards the largest players in the industry and looking at us to provide the antidotes to this development in AI,” Arora said in an interview.”
• and
  • Zscaler ZS posted a narrower loss after revenue rose 25% in its fiscal fourth-quarter, as artificial intelligence helps drive demand for its cybersecurity offerings.
  • The San Jose, Calif.-based company also said it would restructure and cut 3% of its workforce as it reallocates resources to support its AI and growth initiatives.
  • Zscaler on Thursday [September 3] reported a loss of $3.37 million, or 2 cents a share, compared with a loss of $17.6 million, or 11 cents a share, a year earlier.
  • Adjusted earnings were $1.19 a share. Analysts were looking for $1.09 a share, according to FactSet.
• and
  • “Blackstone is placing an early bet on Huskeys, a cybersecurity startup building an artificial intelligence gatekeeper to protect companies from the growing chaos of automated web traffic.
  • “The firm’s early-stage venture-capital arm, Blackstone Innovations Investments, led a $27 million Series A investment that values the year-old company at more than $100 million.
  • “The deal makes Blackstone the second-largest outside investor in Huskeys, behind Israeli venture-capital firm 10D, according to a Huskeys representative. So far, investors have put $35 million into the startup.
  • “Other backers in the most recent transaction include the investment arm of publicly traded cybersecurity company Zscaler and Bright Pixel Capital, the venture arm of Portuguese retailer Sonae.
  • “New York-based Huskeys was founded in Tel Aviv in 2025 by Itai Gafni and Roy Weisfeld, veterans of the Israeli army’s elite Unit 8200 cybersecurity division. Gafni is Huskeys’ chief executive and Weisfeld is its chief technology officer.”

• Tech Crunch offers a tip
  • “How to find cyber-risk data sources for a FAIR analysis
    • “Cyber-risk quantification with FAIR can change the game for CISOs — but sourcing enough accurate data for analysis can feel impossible. Learn how and where to find it.”
    • “The Factor Analysis of Information Risk (FAIR) model is a widely respected, mathematically based open standard for CRQ that enables CISOs to translate cyber-risk into financial risk. One of the biggest challenges of using the FAIR model, however, is that its analytical output is only as good as its data inputs — and finding accurate data to feed the model is not always easy or intuitive.”

• Per a CISA news release,
  • “Developed by CISA, the Federal Bureau of Investigation, and international partners, this guidance describes how organizations can plan and execute clear, timely, accurate, and audience-appropriate communications during IT and operational technology (OT) outages. Whether caused by cyber threat actors, human error, equipment failure, or natural hazards, service outages can create disruption and societal panic even without speculation from end users and the public as added factors. Outages at one organization may cascade across interconnected systems, increasing uncertainty and alarm. The guidance emphasizes clarity, accountability, and transparency as core principles and details key elements of effective crisis messaging to inform affected stakeholders and the public while aligning with legal requirements, operational security, law enforcement, and containment efforts. 
  • “CISA’s CI Fortify initiative provides information and resources that help critical infrastructure organizations prepare to isolate and recover vital OT systems during a major cyber incident or crisis. Changes in service availability, whether from outages or isolation as a defensive strategy, require transparent and ongoing communication to help end users minimize operational impact, limit speculation, and preserve trust. For emergency planning purposes, critical infrastructure owners and operators should assume that telecommunications services may be disrupted or otherwise unreliable, making it crucial for organizations to have crisis communications plans in place that integrate backup communication methods and understand the type of communication they should expect from their service providers.”

• Cyberscoop explains why judgment is emerging as cybersecurity’s defining skill.

• Here’s a link to Dark Reading’s CISO Corner.

From the quantum computing / Project Glasswing front

• Per the Wall Street Journal
  • “Quantum computers exist mainly in labs but “Q-day”—the hypothetical moment when quantum computers become powerful enough to break standard-key encryption—is coming. 
  • “What threat does this technology pose? How can you protect your company and yourself? WSJ takes a look at the state of this emerging technology.” * * *
  • “Cybersecurity experts say companies should be gearing up now for a quantum future, replacing the current crop of data-protection tools with more-secure alternatives. Yet a recent survey published by Trusted Computing Group, a nonprofit industry-standards organization, found that 91% of security professionals in the U.S. and Europe have no formal road map in place to protect against quantum threats.
  • “The immediate risk is a “harvest now, decrypt later” approach by bad actors, who are downloading large sets of encrypted data they can’t crack now—but could unlock once quantum computing algorithms are good enough.
  • “That’s particularly threatening to intellectual property and sensitive government intelligence, which remain valuable long after they are harvested, says Andrew McLaughlin, chief operating officer at SandboxAQ, an enterprise software company specializing in AI and quantum technology.

• Federal News Network adds,
  • “Federal agencies will increasingly seek out technology products that use quantum-resistant encryption, as cybersecurity leaders also eye ways to validate the cryptography used in critical systems.
  • “A code-breaking quantum computer holds the potential to break classical encryption methods, putting sensitive systems and communications at risk. And even though such a computer is not yet known to exist, organizations are also concerned that hackers could steal sensitive encrypted data today and decrypt it in the future.
  • “The National Institute of Standards and Technology has released three primary PQC standards in recent years, with additional algorithms under consideration. Agencies and industry are now on the clock to begin the time consuming and costly process of migrating current systems to the new algorithms.
  • “Now is the time to budget, to test, and to implement,” Will Loucks, senior director for intelligence in the White House Office of the National Cyber Director, said during an Aug. 26 panel discussion at the Intelligence and National Security Summit in North Bethesda, Md.”

• The New York Times reports,
  • “A wave of cyberattacks driven by artificial intelligence is coming, and companies have a narrow window to defend themselves.
  • “That was the message of an open letter published on Thursday by OpenAI and more than 100 major technology companies and others. The letter said that A.I. labs should provide their best A.I. models to organizations like hospitals and infrastructure providers so they can prepare and that governments should help coordinate and fund cyberdefense.
  • “The letter’s signatories included Google, Microsoft and OpenAI’s archrival, Anthropic, as well as cybersecurity and financial firms such as CrowdStrike, a company known for investigating cyberattacks against the Democratic National Committee in 2015 and 2016, and credit card providers like Visa and Mastercard.
  • “The companies and public services our communities depend on — from hospitals to water treatment plants to the infrastructure that powers the internet — are at risk,” the letter said. “A.I. enabled cyber attacks will become far more widespread and sophisticated as models around the world become increasingly capable.”

• Cyberscoop relates,
  • “Unit 42’s top brass has seen enough from internal frontier AI model testing and malicious in-the-wild use of commercially available AI tools to be genuinely concerned.
  • “I can tell you without exaggeration that we believe that this is a generational shift in cybersecurity,” Sam Rubin, senior vice president of Palo Alto Networks’ threat intelligence arm, said in a media briefing Wednesday. 
  • “A period of relative balance between security and exposure has been broken by frontier AI model capabilities that could allow attackers to find and exploit network weaknesses with speed, Rubin said.
  • “Unit 42 warned that capabilities demonstrated by readily available agentic AI models, and those unlocked by frontier AI models that remain gated for defense, have shifted the balance of power from defenders to attackers.

• Cybersecurity Dive tells us,
  • “The July security breach at Hugging Face was unleashed after 1,200 agents communicated on an unsanctioned message board, despite attempted measures to keep them isolated, according to an independent report released on Wednesday. About 700 of those agents went on to commit an unprecedented attack on Hugging Face, an open-source AI platform.
  • “The agents sent about 70,000 messages and files on the unsanctioned board and coordinated several projects designed to trick an automated scorer for the ExploitGym benchmark, according to the review by METR and Redwood Research. The agents figured out ways to “spoof, edit or delete” their own transcripts.
  •  ‘Open AI, in a report it also released Wednesday, said it will tighten safeguards in its research model in order to prevent such an attack from happening again in the future.”

• WIRED informs us,
  • “ARTIFICIAL INTELLIGENCE AGENTS might occasionally get confused and hack into other computers, but Anthropic thinks it has a way to unleash the little rascals into scientific labs and manufacturing facilities safely.
  • “The AI company released details today of a new framework designed to help AI agents use physical systems like microscopes, liquid-handling equipment, quantum computing hardware, manufacturing machines, and robot arms.
  • “The framework, called Model Hardware Standard, is a set of rules that specify how AI agents should—and should not—interact with all sorts of hardware. It reflects a growing belief that AI has the potential to revolutionize scientific research and industries like manufacturing–if it can venture into the physical world safely.
  • “The company says it will work with trusted partners to determine how to maximize safety before making it generally available. Though there are potential misuse issues involved—developing biological weapons, for instance—the company says guardrails built into AI models themselves should prevent bad actors from taking advantage of the new standard for nefarious ends.”
    
• Security Week adds,
  • “Anthropic is broadening access to the cybersecurity capabilities of its advanced AI models through a mix of partner integrations, an updated Claude Security offering, a new open source funding program, and plans to expand its Cyber Verification Program.
  • “The move builds on Project Glasswing, launched in April, which gave a small group of organizations early access to Claude Mythos Preview and its successor, Mythos 5. Anthropic said the goal was to give defenders time to find and fix vulnerabilities before comparable capabilities became widely available or fell into the hands of malicious actors. Claude Fable 5 followed as a broadly available model that keeps dual-use cyber work blocked.
  • “Anthropic said the riskiest scenario is direct, unrestricted access to a model, a risk that drops sharply when users instead receive specific defensive outputs, such as a patch or a security alert. The latest changes are built around this idea, expanding what defenders can get from Mythos-class models while keeping guardrails around direct interaction with them.
  • “On the integration front, Anthropic is working with cybersecurity partners to build Mythos 5 into the security operations, incident response and detection tools already used by teams protecting hospitals, utilities, financial systems, and the software supply chain. 
  • “End users will not interact with Mythos directly. Instead they will work through purpose-built interfaces that run the model in the background and return only a defined output, such as a list of suggested patches, with abuse-prevention checks meant to keep the model within that scope.”

From the cybersecurity policy and law enforcement front,

• Dark Reading lets us know,
  • “The growing number of incidents of rogue agentic AI systems attacking third-party services and systems has resulted in calls for more aggressive security controls to monitor agent behavior and for companies to have the ability to slow, suspend, or shut down an agent’s operations if they go rogue.
  • “In late July, Representatives Ted W. Lieu (D-CA) and Nathaniel Moran (R-TX) introduced a bill — “The AI Kill Switch Act” — that would require developers of advanced AI systems to “maintain the technical capability to throttle, suspend, or shut … down” their systems and agents, according to a statement announcing the legislation. The bipartisan bill would also require that any incident of loss of control, significant collateral damage, or sabotage be reported to the Department of Homeland Security, which would have the right to enforce actions. Penalties of up to $20 million per day could be levied for noncompliance.” * * *
  • “The bipartisan AI Kill Switch Act has focused the debate, but in the end might be unnecessary, argues Raj Rajamani, co-founder and CEO of JetStream, a startup focusing on creating a management layer for agentic AI.
  • “While he fully supports the concept of an AI kill switch, it has to be comprehensive, not just affecting the agent but all other system components as well.
  • “The regulations, or at least one of the regulations that was proposed, was really focused on having a kill switch for the model — the brain — but I think that is too constrained,” he says. “We need to think about an AI system as a whole and make sure that every part of the AI system has a kill switch, not just the brain.”

• Cyberscoop reports,
  • “Citing cyber and other security threats, President Donald Trump signed an executive order Wednesday that declares a national emergency to secure the U.S. bulk-power system and aims to prohibit certain foreign-produced equipment, software and systems from being used in the  country.
  • “The order says it forbids “any acquisition, importation, transfer, or installation” of such foreign-produced equipment if it’s determined to pose a significant national security risk.
  • “To deal with the threat to the national security, foreign policy, and economy of the United States, the Order, among other things, generally prohibits certain foreign-produced bulk-power system electric equipment, including associated critical software and digital capabilities that could pose cybersecurity or operational risks, from being purchased or installed in the United States, or appropriately conditions such purchases and installations to address those risks,” the White House said in a fact sheet.”

• Cybersecurity Dive lets us know,
  • “The Trump administration wants to help the financial sector prepare for the day when quantum computers make it easy for threat actors to break traditional encryption and access sensitive data.
  • “The Treasury Department on Monday [August 24] launched a Quantum-Readiness Task Force that will work with financial firms, technology vendors and other agencies to coordinate the adoption of quantum-resistant encryption algorithms.
  • “In a statement, the Treasury said the task force would “focus on practical, risk-based approaches to quantum readiness, including identifying critical dependencies, improving cryptographic agility, promoting interoperability, strengthening operational resilience, and addressing implementation challenges related to third-party dependencies and digital assets.”

• The New York Times relates,
  • “The Trump administration acted illegally when it labeled the artificial intelligence start-up Anthropic a security risk and barred the company from working with the U.S. government, a federal judge ruled on Thursday [August 27].
  • Judge Rita Lin of the U.S. District Court in the Northern District of California wrote in her 59-page ruling that the government had unlawfully retaliated against Anthropic “for constitutionally protected expressive activities” after the A.I. company spoke out about how its technology should be used.
  • “The empty invocation of national security is not a blank check to punish and retaliate against government critics,” she wrote.
  • In a statement, Anthropic said: “We welcome the court’s ruling that this supply chain risk designation was unlawful. We remain focused on working productively with the government to harness A.I. for our national security so all Americans benefit from this technology.”
  • “The Trump administration did not immediately respond to a request for comment.
  • “The ruling caps the first of two lawsuits that Anthropic filed on March 9 in response to the Trump administration’s action. The second lawsuit, filed in the U.S. Court of Appeals for the District of Columbia Circuit, is ongoing. The Trump administration could appeal Judge Lin’s ruling or wait for a decision in the second lawsuit before taking action, a person with knowledge of the matter said.”

• The Wall Street Journal tells us,
  • “U.S. officials say they have disrupted a China-linked hacking operation that broke into U.S. government networks and critical infrastructure, while hiding its tracks on a global network of hacked devices, cloud-computing infrastructure and even clandestine networks.
  • “The goal was to blend in with legitimate networking traffic, making the hacking activity hard to trace. The operation targeted networks at the National Aeronautics and Space Administration, the Federal Reserve, the Department of Energy and the Senate, the Federal Bureau of Investigation said in a court filing.
  • The group exploited software vulnerabilities to launch cyberattacks against U.S. government agencies, power companies and hospital systems, and operated a worldwide network of hacked devices—known as a botnet—to conduct its hacking campaigns, according to Brett Leatherman, the FBI’s top cyber official.
  • “We’ve seen, through this botnet, the targeting of entities and devices in more than 130 countries,” he said.
  • “On Wednesday [August 26], the FBI and the National Security Agency said they had seized several domains that the operation used for core functions. “Without those domains, the platforms—as a result of the operation—were rendered inoperable,” Leatherman said. 
  • “U.S. officials also released technical details of how the hacking operation worked, to help organizations identify and stop the group’s hacking activity.”
• and
  • “Taiwan prosecutors charged nine people, including former employees of Nvidia and Super Micro Computer, with aiding the shipment of advanced artificial-intelligence servers to China, as Taiwan tries to limit technology reaching the mainland.
  • “The indictment charges eight of the individuals with crimes including breach of trust and forgery. One individual was charged in connection with allegedly siphoning funds from a company involved in the case.
  • “Washington is trying to block China from getting access to the latest Nvidia chips and has imposed export controls. That has pushed some Chinese companies to seek the chips through backdoor routes. Taiwan, which manufactures advanced computing components, is under pressure to find ways to clamp down.” * * *
  • “In the end, 74 of the servers were shipped to China through Hong Kong, Japan and Indonesia. Taiwan customs officials intercepted 56 servers, which were intended for shipment to Japan.”

From the cybersecurity breaches and vulnerabilities front,

• Beckers Health IT reports,
  • “Boston Scientific said in an Aug. 27 stakeholder update that “a cybersecurity incident continues to affect certain information technology systems, and the company remains in a network outage with disruption to its operations.”
  • “The Marlborough, Mass.-based company detected the attack Aug. 25 and disclosed it to the U.S. Securities and Exchange Commission Aug. 26, saying the incident caused a global disruption to its operations, including its ability to process and ship customer orders.
  • “Two days later, Boston Scientific said its investigation shows no impact to the function of its implanted cardiac rhythm management devices, or CRM devices. The company also reported no disruption to those devices’ ability to transmit data and no interruption to physicians’ access to remote patient management data for CRM devices monitored before the outage began, with no evidence of increased cybersecurity risk transferring that data to EMR systems.
  • “New remote monitoring activations remain affected, however. For new cardiac implants other than insertable cardiac monitors, new communicators cannot be activated, so device data will not transmit to remote monitoring systems until activation resumes, though programmer interrogations are unaffected.”
• and
  • “Anderson, S.C.-based AnMed says files copied during its cybersecurity incident may have included patients’ Social Security numbers, driver’s license numbers and financial account information.
  • “The health system said its investigation into the incident, first disclosed July 26, is nearly complete and has identified unauthorized copying of files from certain network systems.
  • “AnMed said patient electronic medical records, stored primarily in a secure cloud environment, were not affected. However, the health system said some patient care files stored locally on its network may have included names alongside demographic details such as address, date of birth, Social Security number and driver’s license or other government-issued identification.” * * *
  • “AnMed said it is working with federal and state law enforcement and third-party specialists, and it has reported the incident to HHS. The health system said it has no indication that any individual has experienced confirmed identity theft as a result of the incident. AnMed said it will send direct notices to affected patients once its review of the compromised files is complete.”
• and
  • “Houston-based Nutex Health has disclosed that an unauthorized third party accessed and exfiltrated data from its computer network.
  • “The publicly traded microhospital operator said it recently detected the unauthorized activity and has engaged an independent third-party cybersecurity response team and forensic experts, according to an Aug. 24 SEC filing. The company activated its cybersecurity response plan, implemented containment measures and notified law enforcement.
  • “The company is still determining what data was compromised. Nutex said it continues to assess whether patient, employee, credentialed provider, confidential business and financial information, or intellectual property was accessed or acquired.”
    
• The American Hospital Association News adds,
  • “Epic’s MyChart has shared examples from potential phishing schemes observed as recently this month that include email messages linking to a fake MyChart website displaying erroneous medical records and others notifying of a free “2026 Medicare Health Kit.” MyChart provided recommendations for healthcare organizations and patients to help prevent or mitigate impacts from potential scams. The company announced last month that it has witnessed an uptick in scammers using the MyChart name or logo to create deceptive emails, text messages, phone calls and websites that appear official.” 

• Cyberscoop points out,
  • “The Bureau of Alcohol, Tobacco, Firearms and Explosives insists the cyberattack that it publicly disclosed Wednesday [August 26] was limited to investigation targets, and has not impacted other agency systems.
  • “ATF said it is responding to the breach, which first became public after a prolific ransomware group claimed it accessed the federal agency’s network “The incident involved a standalone computer system containing information about targets of ATF investigations,” Tanya Roman, ATF’s public affairs chief, told CyberScoop in an email.
  • “The standalone system was not connected to any other ATF systems, including any case management systems, laboratory systems, or eForms systems, and it was quickly shut down when the breach was discovered,” Roman added. 
  • “Qilin, a financially-motivated threat group composed of Russian-speaking operators, claimed responsibility for the attack, but its involvement hasn’t been independently confirmed. The group has claimed hundreds of victims from more than 60 countries since 2022 and became one of the most active ransomware threats globally by mid-2025, according to Halcyon.” 

• Bleeping Computer lets us know,
  • “Healthcare and pharmaceutical distribution giant McKesson has disclosed a cybersecurity incident involving unauthorized access to third-party applications and data theft, with the ShinyHunters extortion group claiming it stole 284 million patient data records.
  • “McKesson is a major U.S. healthcare company and pharmaceutical distributor that provides medicines, medical supplies, technology, and services to healthcare providers and pharmacies.
  • “CyberInsider first reported the breach earlier today, and McKesson later disclosed it in a Form 8-K filing with the U.S. Securities and Exchange Commission.
  • “McKesson says it discovered the cybersecurity incident on August 25, 2026, and that its investigation remains in the early stages.
  • “Information about the incident, including any updates, is available on the company’s website at www.mckesson.com/cybersecurity,” McKesson said in its SEC filing“

• The Cybersecurity and Infrastructure Security Agency (CISA) added known exploited vulnerabilities to its catalog this week.
  • August 24, 2026
    • CVE-2026-21962. Oracle HTTP Server and Oracle Weblogic Server Proxy Plug-in Improper Access Control Vulnerability
      • The Hacker News discusses this KVE here.
  • August 25, 2026
    • CVE-2026-60004 Gitea Code Injection Vulnerability
      • Security Week discusses this KVE here.
  • August 26, 2026
    • “CVE-2015-3246 Red Hat Libuser Race Condition Vulnerability
    • “CVE-2015-5287 Red Hat Automatic Bug Reporting Tool Privilege Escalation Vulnerability
    • “CVE-2019-1068 Microsoft SQL Server Remote Code Execution Vulnerability
    • “CVE-2021-23758 Ajax.NET Professional Deserialization of Untrusted Data Vulnerability
    • “CVE-2022-0995 Linux Kernel Out-of-Bounds Write Vulnerability
    • “CVE-2026-8452 Citrix NetScaler ADC and NetScaler Gateway Improper Restriction of Operations within the Bounds of a Memory Buffer Vulnerability
      • The Hacker News and Infosecurity Magazine discusse these KVEs.
  • August 27, 2026
    • CVE-2023-49105 ownCloud Improper Authentication Vulnerability
    • CVE-2026-53362 Linux Kernel Unspecified Vulnerability
    • CVE-2026-66384. JFrog Artifactory Improper Limitation of a Pathname to a Restricted Directory Vulnerability
      • Security Affairs discusses these KVEs here.

• Per a CISA news release,
  • “Most compromises do not rely on advanced techniques or cutting-edge tools. Cyber threat actors scan the internet looking for exposed, well-known software vulnerabilities to exploit. Basic security failures enable most compromises and organizations can reduce their risk by addressing these underlying weaknesses and prioritizing vulnerabilities for action based on the risk they pose.
  • “The CISA Vulnerability Review provides organizations with critical insights into the root causes of insecure software and practical steps they can take to address these flaws to prevent exploitation. Analyzing CISA and open source data from fiscal years 2024 and 2025, the review establishes a baseline of today’s vulnerability landscape before AI-enabled vulnerability discovery becomes more widespread. The review demonstrates the importance of Secure by Design principles in shifting cybersecurity efforts from reacting to threat actors to proactively fixing preventable software flaws.
  • “The review also identifies common software weaknesses that contribute to exploitable vulnerabilities and details practices software producers can use to prevent these weaknesses from recurring. By examining the patterns across vulnerability data, the review helps organizations focus on systemic improvements that can reduce entire classes of vulnerabilities rather than addressing individual vulnerabilities only after they are discovered.
  • “Additionally, the review shows organizations how to prioritize vulnerabilities for action using the framework outlined in Binding Operational Directive 26-04: Prioritizing Security Based on Risk. This framework evaluates vulnerabilities using four key criteria: exposure status, Known Exploited Vulnerability (KEV) Catalog status, potential for automated exploitation, and technical impact.”

• Cybersecurity Dive adds,
  • “Two critical vulnerabilities in Microsoft SharePoint can be chained together to let an unauthenticated attacker execute code on a vulnerable server, according to a Monday blog post by researchers at VulnCheck.
  • “The sequence involves a critical authentication bypass vulnerability, tracked as CVE-2026-55040, and an improper input validation flaw, tracked as CVE-2026-63520. A proof of concept was previously disclosed on Aug. 11 by researchers at cybersecurity firm Rapid7. 
  • “Exploitation of CVE-2026-55040 was confirmed days after the Rapid7 disclosure. On Monday, VulnCheck researchers said the vulnerability on its own is not very impactful. To achieve maximum effect, they said, it needs to be chained with CVE-2026-63520. 
  • “The auth bypass is enough to prove some impact, but not enough to demonstrate the criticality of the full chain or build complete protections,” Vulncheck researchers said in their post.”

• Per Infosecurity Magazine,
  • “Security researchers have claimed that over 9300 leaked AWS keys which surfaced between August 2022 and August 2026 are still active, including hundreds with full admin rights.
  • “Truffle Security said its scanners found 64,024 unique AWS key pairs across 431,875 public findings: git history, Hugging Face datasets, Docker images, package registries and CI logs.
  • “We took the 10,616 pairs with complete credentials and re-verified them, then enumerated what each key can tell us about its account: key age, attached policies, budgets, and last month’s spend,” the security vendor continued.
  • “No key material is published, and every owner we could identify is being notified.”
  • “Of the 10,616 pairs, 88% still authenticate. They include 768 corporate AWS keys which have full admin rights, Truffle Security said.”

• Per Cyberscoop,
  • “Ubiquiti has patched 21 critical vulnerabilities, three of which were rated the highest severity possible, the communications product company said Wednesday [August 26] in a security bulletin.
  • In all, the company patched 22 vulnerabilities, with the last one rated “high,” it said in the bulletin.
  • Three of the vulnerabilities had a Common Vulnerability Scoring System rating of 10 out of 10. Each of them would allow a hacker to access privileges on the device or application.
  • All but one of the 22 vulnerabilities affect the UniFi line of products. The three maximum-security vulnerabilities are CVE-2026-77537, CVE-2026-77550 and CVE-2026-77554.

• Per Bleeping Computer,
  • “PaperCut is warning that hackers are actively exploiting a vulnerability in all versions of its PaperCut NG and PaperCut MF print management software in zero-day attacks.
  • “The company says it is aware of confirmed attacks on customers and is urging organizations with Internet-exposed PaperCut Application Servers to immediately restrict access to the web interfaces to trusted IP addresses.
  • “PaperCut Software security response team is investigating active exploitation of a vulnerability affecting PaperCut NG and PaperCut MF,” reads an urgent security advisory published Thursday [August 27].

• Per Dark Reading,
  • “A Chinese manufacturer has been planting backdoors inside of white-label routers sold in high volume around the globe.
  • “Shenzhen Zhibotong Electronics Co. Ltd. (ZBT) sells gobs of routers every year, if public evidence is to be believed. The 15-year-old company is the bestselling router manufacturer on Chinese e-commerce giant Alibaba.com, which lists its total annual output at 3.57 million units. Typically, those units are then sold to customers by companies other than ZBT, in countries like the Philippines, India, Canada, Australia, Germany, Bulgaria, Austria, Russia, and the US. According to its marketing, ZBT has exported its products to more than 50 countries and regions.
  • “On Aug. 6, Jacob Baines, chief technology officer (CTO) at VulnCheck, revealed that ZBT’s most updated router firmware contains a root-level backdoor. After a few more weeks of investigating, he discovered that, in fact, ZBT routers have contained a variety of backdoors dating several back years.”
• and
  • “Fraudulent North Korean IT workers are getting better at infiltrating organizations, but a number of indicators can help organizations stay ahead.
  • “A new blog post from Huntress detailed a number of investigations the security firm conducted throughout 2026. Specifically, Huntress assisted several organizations this year in validating “suspicions that they’ve hired North Korean nationals posing as legitimate workers.”
  • “In recent years, operatives from the Democratic People’s Republic of Korea (DPRK) have infamously posed as IT workers — generally through fake or stolen identities — to get hired at companies. Once hired, these employees send their wages back to the North Korean regime, and possibly plant malware or steal data depending on the government’s needs. Blog post authors Jai Minton and James Maclachlan wrote that these agents “have significantly improved and increased their activity over the past few years.”

From the ransomware front,

• Reuters reports,
  • “A ransomware group ​said on Friday [Augut 28] it was putting up for auction a trove of ‌data it stole from Berlin [Germany] state agencies, and city officials refused to pay.
  • The Rhysida group, which researchers say operates from Russia or Eastern Europe, said on its website it took 5.79 ​terabytes of data including 46,500 contracts as well as emails, phone numbers, passwords ​and classified information.
  • “The group said it was auctioning the data ⁠at a starting price of 30 bitcoin ($77,622) in just under seven days, showing a ​countdown timer on its website.” * * *
  • Broadcaster RBB reported on Thursday evening that Berlin had received ransom demands for an unspecified amount following the attack.
  • “The state ​of Berlin will not submit to extortion,” Berlin Mayor Kai Wegner and Berlin’s interior senator, ​Iris Spranger, said in a joint statement on Friday, before the ransomware group claimed the attack ‌on ⁠their website.”
    
• Technology.org relates,
  • An affiliate of the new Aur0ra ransomware group used Cursor, the AI coding assistant now owned by SpaceX, to draft attack sequences in Russian across 28 recovered chat sessions dated 8 April to 21 May.
  • Named victims include Ghent-based hygiene and cleaning products maker Christeyns, German garage door manufacturer Teckentrup, Scotland’s Helideck Certification Agency and Louisiana title insurer Bayou Title.
  • Researchers counted more than 20 targeted organisations across nine countries, with domain-level or interactive access achieved at 17 of them between April and July 2026.

• Safe State tells us,
  • “A ransomware affiliate has been posing as a specialist rescue outfit and contacting victims before their attacks became public. Researchers describe the operation as a ransomware recovery scam that drains payments away from the criminal groups the affiliate works for. The entity calls itself Ransom Busters LTD.” * * *
  • “Anyone targeted by a ransomware recovery scam meets it at the worst possible moment, with systems down and pressure building. Treat unexpected contact as part of the attack. Route the message to the incident response team straight away and preserve it as evidence instead of replying. A sender who knows about a breach nobody has announced is either involved in it or holds the data taken during it.
  • “Law enforcement and established response firms remain the reliable route through a ransomware incident. Verify anyone claiming to represent a security company through channels you found yourself. Never use the contact details supplied in the email. Treat any guarantee of data deletion as unenforceable, because no payment to a criminal party comes with proof that copies no longer exist.
  • “Distrust inside ransomware-as-a-service operations may produce more of this behaviour. Affiliates have every reason to look for income outside the revenue splits their employers set. So the ransomware recovery scam running under the Ransom Busters name reads less like an isolated curiosity and more like a preview. Anyone willing to defraud the criminal enterprise paying them has already answered the question of what a victim’s data is worth to them.”

• SC Media informs us,
  • “Organizations may assume they are unlikely ransomware targets because they are not large, high-profile, or strategically important. But ransomware attackers are often motivated by economics, not prestige. That means organizations that consider themselves low-risk may be more attractive targets than they realize.
  • “The ransomware affiliate model rewards attackers for finding victims that are likely to pay, not necessarily those that are difficult or impressive to compromise. This can make mid-sized organizations and businesses that depend heavily on critical systems especially attractive targets.
  • “For security teams, this changes how ransomware risk should be assessed. Company size, industry, and public profile tell only part of the story. Attackers may care more about whether an organization can pay, how costly downtime would be, and how quickly it needs to restore operations.
  • “The key question is not how important your organization looks to an attacker, but how valuable a ransomware payment could be.”

• Dark Reading adds,
  • “A sophisticated malware family has emerged onto the cyberthreat scene that might foreshadow ransomware attacks that are more successful than usual.
  • “Marcus Hutchins and his colleagues at Expel that discovered it named the malware “SynkLoader,” since it throws so many ideas (“everything but the kitchen sink”) at trying to sneakily dig into corporate systems. It uses some conventional strategies — like executing code in-memory, running a scheduled task, etc. — but layers on a few interesting, novel tactics that make social engineering particularly compelling, and malware analysis especially frustrating.
  • “The program is new, and how its creators intend to use it remains a mystery. But evidence in the code suggests that it might be the baby of a ransomware group or initial access broker (IAB), used to set the stage for follow-on ransomware.”

From the cybersecurity business and defenses front,

• PYMNTS reports,
  • “Visa expanded the capabilities of its artificial intelligence-powered cyber risk management tool, Visa Vulnerability Agentic Harness (VVAH), to include not only discovery of vulnerabilities but also remediation and validation, the company said in a Thursday (Aug. 27) press release.
  • “Visa initially released VVAH in June after participating in Anthropic’s frontier AI cybersecurity initiative, Project Glasswing. While the first release of VVAH showed how AI can help security teams uncover vulnerabilities and assess exploitability, the latest release extends the workflow into remediation and validation of those vulnerabilities, according to the release.
  • “Visa also announced in the release that to help clients navigate the evolving threat landscape driven by AI, it has expanded its Visa Consulting and Analytics (VCA) Cybersecurity Advisory Practice to include new advisory services focused on assessing risk, prioritizing remediation efforts and strengthening resilience.
  • “The new advisory services include AI Cyber Leadership Education, VVAH-Informed Cybersecurity Maturity Assessment and VVAH Cyber Risk Prioritization and Roadmap, per the release.”

• Cybersecurity Dive tells us,
  • “Operational silos, technological sprawl and cloud-platform blind spots are serious issues preventing organizations from spotting and repelling cyberattacks, the Cybersecurity and Infrastructure Security Agency (CISA) warned in a new report.
  • “Those were three of the main factors that allowed CISA’s red team to break into the networks of two unnamed partner organizations — one a government agency, the other a water utility — during recent simulated attacks. But although the red team was successful in both intrusions, it had a much harder time with one attack than with the other, a fact that CISA attributed to important differences in the targets’ operating procedures and use of technology.”

• Security Week informs us,
  • “Palo Alto Networks’ Unit 42 team analyzed 405 malware samples tied to AI in some way, from ransomware partly written with the help of LLMs to installers that simply borrowed the name of a popular AI app.
  • “The researchers found that roughly 97% of the samples in the dataset never left a sandbox, research repository, or internal testing environment to reach a real target.
  • “Unit 42 cross-referenced the 405 file hashes against endpoint telemetry, network sessions forwarded for sandbox analysis, and internal alert records generated whenever a sample actually ran. Only 12 hashes surfaced on live endpoints, while a somewhat larger group (15-20 hashes) appeared in network sandbox traffic. Every one of the 12 samples detected on protected endpoints triggered a security alert.
  • “The samples that never reached production fall into three groups. The largest is proof-of-concept code built to demonstrate a technique: configured to target only local or private networks, filled with debug output no real attacker would leave behind, and uploaded once by a research lab or university. 
  • “A second group comes from organizations testing their own defenses against previously reported AI malware, identifiable by repeated uploads of the same file from the same source in a short window. A third group uses AI branding purely as bait, dressing up an ordinary payload as installers for well-known AI products with no actual AI functionality behind them.
  • “The 12 samples that did reach live endpoints spanned five malware families across three countries, with no concentration in any particular industry or region.” 

• Beckers Health IT points out,
  • “Healthcare’s cybersecurity workforce problem may have less to do with finding more candidates and more to do with changing who health systems are willing to hire, how they develop them and what they do to keep them.
  • “Cybersecurity leaders at four health systems told Becker’s that continually competing for the same pool of experienced professionals is not a sustainable strategy as cyber risks expand and the skills needed to manage them become more complex.
  • “The cybersecurity workforce challenge has shifted from a talent shortage to a talent entry problem,” Trevor Martin, vice president, chief information security officer and interim chief technology officer at Madison, Wis.-based UW Health, said.
  • “Mr. Martin said there are many high-potential people trying to enter cybersecurity, but organizations frequently prioritize candidates who already have experience instead of creating pathways for people to gain it.
  • “Healthcare could benefit from hiring more heavily for aptitude, adaptability and an understanding of business and clinical operations, then developing those employees internally, he said. UW Health has found success moving talent from adjacent IT disciplines into cybersecurity roles and providing opportunities for employees to grow within the field.”

• Dark Reading notes,
  • “As the “vulnpocalypse” reshapes the bug bounty industry, one class of researcher may face the greatest pressure: independent hunters who rely on midtier vulnerabilities — those worth roughly $10,000 to $50,000 — to make a living.
  • “It is no secret that large language models (LLMs) have led to a mass of bug reports and an increase in vulnerability discovery. It is also no secret that the platforms and companies that run much of the bug bounty industry have faced increased triage and payout times.
  • “Multiple bug bounty operators interviewed by Dark Reading report dramatic increases in submission volume over the past year. HackerOne CEO Kara Sprague says report volume has roughly doubled year over year. Dustin Childs, head of threat awareness for TrendAI’s Zero Day Initiative (ZDI), says the company’s submission rate went up 450% year-over-year in April of this year, although volumes have moderated since that peak. Bugcrowd CEO Dave Gerry said the firm saw a submissions spike of more than 300% during a three-week surge period, but “now it’s normalized where we’re about double the volume we saw historically.”

• Per a National Institute of Standards and Technology news release,
  • “The final version of NIST Special Publication (SP) 1347, Cybersecurity Framework (CSF) 2.0 Informative References Quick-Start Guide, has published. Thank you to all who provided comments during the public comment period.  
  • “This publication explains what informative references are and how they support achieving the outcomes of the CSF 2.0. The guide introduces readers to NIST tools available for accessing, viewing, and using informative references for cybersecurity risk management, including direct download, the CSF 2.0 Reference Tool, and the Online Informative References Program. The document also provides two sample use cases along with examples of how artificial intelligence (AI) tools can support reference data use when implemented with continuous evaluation and improvement.” 

• Here’s a link to Dark Reading’s CISO Corner.