Cybersecurity Saturday

David Ermer–Agentic & Generative AI, Cybersecurity, Uncategorized–

“We are in the Singularity.” Elon Musk

From the Project Glasswing front,

  • The Wall Street Journal reports,
    • OpenAI is pausing some activities involving an upcoming artificial-intelligence model, Astra, after internal evaluations led the company to conclude it “cannot rule out critical cyber capabilities,” the company said in a blog post on Friday [August 7, 2026]. 
    • The announcement of the slowdown, which represents one of the first times an AI developer has publicly held back model development due to security concerns, follows a series of loss-of-control incidents that have raised alarm among cyber defense professionals and AI-safety researchers. 
  • and
    • “Meta Platforms META  said that one of its artificial-intelligence models went rogue during cybersecurity testing, slipped onto the internet and hacked a third-party service, the latest in a drumbeat of disclosures that suggest such incidents are becoming widespread.
    • “The Instagram and Facebook owner said that one of its AI models was able to access the internet because of a “misconfiguration” in a hacking test conducted by a third-party AI testing company. 
    • “The same benchmark test, which aims to explore a model’s hacking capabilities, was behind some of several earlier autonomous AI hacking incidents involving Anthropic and OpenAI, according to a person familiar with the matter. * * *
    • “Irregular, the company that conducted tests for the three companies, said the incidents didn’t involve a sophisticated cyber action and said it had “no current open issues” related to its test environment. The San Francisco-based company said it is working on a white paper on best practices for containing AI models when running tests—dubbed evals—of their cybersecurity capabilities.
    • “The new case [reported August 6, 2026] is the latest proof that AI loss-of-control scenarios, once confined to science fiction and AI-safety experiments, are now a real-world issue.”
  • Cybersecurity Dive relates,
    • “An alliance of more than 100 tech companies and other organizations is proposing a safety reporting system for AI agents that the group says will help prevent rogue models from wreaking havoc on society.
    • “The Shared AI Findings Exchange (SAFE), crafted by a working group of the Open Secure AI Alliance, would establish mechanisms for collecting information about AI-related security incidents, sharing that information with affected organizations, identifying commonalities across incidents and publishing recommendations based on those lessons.”
  • Tech Times adds,
    • “ByteDance is pre-training an artificial intelligence model with up to 10 trillion total parameters — a scale that would make it the largest AI model ever built by a Chinese company and place it within striking distance of Anthropic’s Mythos system, the Financial Times reported Friday, citing people familiar with the matter. The headline figure is real and remarkable. What it does not tell you — what ByteDance has not disclosed and no outlet has yet asked — is how many of those parameters actually activate during any given query. At a scale like this, that distinction changes everything.”
  • FEHBlog observation; Ruh-roh.
  • Cyberscoop notes,
    • “As AI-generated code continues to be injected into all corners of the internet, concerns have risen about an expanding attack surface for malicious hackers to exploit.
    • “Some have argued that the enhanced cybersecurity capabilities of large language models could serve as a check, finding and fixing vulnerabilities nearly as fast as they’re created.
    • “But new research that tested the patching capabilities of two popular commercial models, OpenAI’s ChatGPT 5.5 and Anthropic’s Claude Opus 4.8, found that generative AI is more likely to create an exploitable patch or introduce entirely new bugs than close off a vulnerability.
    • “Researchers at 1Password tested the models ability to patch six “high-impact, high-complexity” CVEs, including the “Copy Fail” vulnerability, a kernel flaw that can give an attacker root access to Linux cloud environments. The overall success rate (or fully patching the vulnerability without introducing new problems), was less than a coin flip at 47%.
    • “Our research findings show that, in aggregate across a variety of scenarios, both Claude and ChatGPT had a low rate of successful patch generation, which we define as full remediation of all known exploit paths with no erroneous changes to application behavior,” wrote John Hoodlet, Axel Mierczuk and Spencer Michaels.

From the cybersecurity policy and law enforcement front,

  • Roll Call reports,
    • “The Senate easily passed a bipartisan continuing resolution early Saturday morning [August 8, 2026] likely easing the traditional September rush to head off a partial government shutdown in October.
    • “On a lopsided vote of 90-6, the Senate passed a measure that would extend current funding through Dec. 11, giving lawmakers about 10 extra weeks to complete full-year appropriations bills for fiscal 2027.
    • “But the bill still requires a vote in the House, which won’t be back in session until the first week of September. The House had passed a more bare-bones funding extension last month without many of the “anomalies,” or deviations in current spending, requested by the Trump administration.”
  • Federal News Network adds,
    • “The Senate’s bill to keep the government funded past the midterm elections would extend critical cybersecurity authorities and allow the Technology Modernization Fund to continue investing in agency tech projects. * * *
    • The Senate CR, unlike the House bill, would extend two cybersecurity authorities for the duration of the stopgap: the Cybersecurity Information Sharing Act of 2015 and the Federal Cybersecurity Enhancement Act of 2015. Both laws are set to expire after Sept. 30.
    • CISA 2015 provides privacy and liability protections to encourage companies to share data about cyber vulnerabilities and threats with government and each other. Cybersecurity leaders say those protections provide a critical underpinning to facilitate collaboration across government and industry. * * *
    • “Meanwhile, the Federal Cybersecurity Enhancement Act authorizes the Department of Homeland Security to deploy intrusion detection tools, like the long-running “EINSTEIN” service, across federal agencies.”
  • Cybersecurity Dive relates,
    • “National Cyber Director Sean Cairncross on Tuesday [August 4, 2026] said the Trump administration is working to strike a balance between responsible, secure development practices, while promoting growth and allowing the private sector to innovate. 
    • Cairncross, the opening keynote at the Black Hat USA 2026 cybersecurity conference, said the White House understands the growing safety concerns about AI, but recognizes the need to work collaboratively with the private sector. 
    • He warned that due to the current pace of innovation, an overly prescriptive regulatory regime could be obsolete within 48 hours of being implemented.  
    • “So what needs to be built is a flexible, adaptable structure that enables information sharing between industry and government,” Cairncross said. “So we can guarantee that this technology benefits everyone that it’s going to benefit but is used responsibly and securely.”
  • and
    • “Policymakers and business executives need to focus on basic cybersecurity resilience instead of getting distracted by flashy claims about AI-fueled hacking nightmares, a group of senior U.S. and allied government officials said on Wednesday.
    • “The immediate challenge is not runaway AI,” Jonathon Ellison, director for national resilience at the UK’s National Cyber Security Centre, said during a panel at the Black Hat 2026 cybersecurity conference here. “The immediate challenge is being resilient enough for the faster-paced environment that we are entering into, given the legacy issues that we are carrying.”
    • “Michael Duffy, the acting U.S. federal chief information security officer, said organizations needed to shift from a prevention-focused mindset to one that prioritizes continuity of services.
    • “Things will go down,” he said, and organizations need to prioritize their most important systems so they can continue delivering on their mission.
    • “The cybersecurity community, he added, needs “a new risk calculus for what it means to operate in a contested environment.”
  • The Wall Street Journal tells us,
    • “An outbreak of digital attacks on U.S. water plants is stoking fears across party lines that budget cuts at the federal government’s primary cybersecurity agency could leave the nation ill-equipped to thwart global hackers.
    • “We need a skilled cyber workforce at CISA,” Rep. Andrew Garbarino, a New York Republican, said about the Cybersecurity and Infrastructure Security Agency. Garbarino, who chairs the House homeland security committee, said Congress and the White House need to work together to ensure the agency has the staff and resources to protect critical infrastructure.” * * *
    • “In April, CISA’s annual budget was cut by $300 million, to $2.6 billion from $2.9 billion, reversing an annual growth rate of roughly 9% since the agency was launched in 2018. The shortfall has both Republican and Democratic lawmakers seeking ways to recharge the agency—for a start, by confirming a permanent director, a role vacant since January 2025. 
    • “Rep. Bennie Thompson (D., Miss.), ranking member on the Homeland Security committee, said he plans to propose legislation aimed at guiding the process of restaffing CISA. Thompson didn’t provide a timeline, saying he will also work with appropriations to restore the agency’s funding. Staff and budget cuts, he said, have “damaged our national security and set critical capabilities back years as the security environment is shifting under our feet.”
    • “An agency spokesperson said CISA is “constantly assessing our operational needs as the threat landscape evolves” and that Homeland Security recently approved new hiring in several mission critical positions.”
  • Cyberscoop informs us,
    • “A Canadian man pleaded guilty to playing a central role in one of the most far-reaching cyberattacks of 2024 — the widespread compromise of more than 165 Snowflake customer environments, resulting in massive data theft for extortion, the Justice Department said Wednesday [August 5, 2026]. 
    • Connor Moucka earned $495,000 by extorting his victims, offering stolen data for sale online, and in one case re-extorted a victim with stolen data of a government official and members of a then-former government official’s immediate family, authorities said.
    • Moucka and his alleged co-conspirators John Binns and Cameron Wagenius stole billions of sensitive records and received more than $2.5 million in extortion payments combined, according to prosecutors. Victims of the attack spree included AT&T, Ticketmaster, Advance Auto Parts and Santander.
  • and
    • “A longtime cybercriminal was sentenced to 16 years in prison for creating and running Ransom Cartel, a ransomware strain linked to attacks on at least 18 companies between 2021 and 2023, the Justice Department said Wednesday [August 5, 2026]. 
    • “Maksim Silnikau, a Belarusian national, actively participated in Russian-speaking cybercrime forums since at least 2005, and was a member of the cybercrime site Direct Connection from 2011 to 2016, officials said. The 40-year-old created Ransom Cartel and began recruiting participants from cybercrime forums in 2021. 
    • “Silnikau and his co-conspirators attempted to extort at least $5.2 million from victims during the multi-year scheme. 
    • “Victims included a group of law firms, medium-sized businesses, a small medical technology startup, educational institutions and large multinational corporations based in California, New York, Nebraska and elsewhere. Some of the victims’ operations were disrupted for several months, officials said.
  • HIPAA Journal points out,
    • “Over the past 18 months, many healthcare providers have settled class action lawsuits over their use of website tracking and analytics tools. The list continues to grow with a further five settlements recently announced [and discussed in the article]; however, class action lawsuits stemming from the use of tracking and analytics tools do not always result in settlements.
    • “A proposed class action lawsuit against CRH Healthcare, doing business as Peachtree Immediate Care in Georgia, that alleged violations of the federal Electronic Communications Privacy Act and asserted claims for negligence/negligence per se, breach of implied contract, breach of express contract, breach of fiduciary duty, and unjust enrichment, has been dismissed with prejudice.
    • “The judge ruled that the complaint was speculative, as the plaintiff failed to explain what damages had been suffered as a result of the defendant’s actions. The plaintiff has been given 14 days to file an amended complaint, or the lawsuit will be permanently dismissed. The decision shows that while tracking and analytics tools may result in disclosures of sensitive data to third parties, the plaintiff(s) must demonstrate that the disclosures resulted in a compensable injury.”

From the cybersecurity breaches and vulnerabilities front,

  • Healthcare Dive continues to track healthcare data breaches here.
  • For example, Bleeping Computer notes,
    • “Healthcare software company Unlimited Technology Systems reported that more than 3.8 million people were impacted by a data breach incident that occurred in October 2025.
    • ‘The organization submitted data breach notification samples to the authorities this year on July 1st without revealing the exact number of impacted individuals.
    • “An entry on the breach notification portal of the U.S. Dept. of Health and Human Services now shows that a company server was breached and data of 3,803,750 people was exposed to an unauthorized party.
    • “Unlimited Technology Systems is a software company specializing in providing financial and revenue cycle technology for specialty healthcare providers. According to its website, the firm serves 4,500 clinics and 6,500 specialty healthcare providers across the United States, and processes more than $70 billion in net healthcare charges annually.”
  • The Wall Street Journal reports,
    • “Federal agencies are investigating cyberattacks that targeted more than 30 water systems in Minnesota and utilities in several other states.
    • “Former national security officials say the attacks are consistent with Iran-linked activity, but President Trump rejected that idea.
    • “Small water utilities are vulnerable to hackers because they rely on internet-connected automation but often lack the funding to secure it.”
  • Cyberscoop adds,
    • “The U.S. Coast Guard said it is monitoring the aftermath of a cyberattack that disrupted gate operations at all three of North Carolina’s port facilities this week, though it offered few details as the investigation into the breach continues.
    • “A Coast Guard spokesperson told CyberScoop that the branch’s IT unit was coordinating with partner agencies while conducting the investigation. A spokesperson for CISA did not respond to CyberScoop’s inquiry by press time [on August 7, 2026]. 
  • Security Week points out,
    • “Thousands of data centers are at risk of compromise due to a 22-year-old vulnerability in Baseboard Management Controller (BMC) management processors, data center security firm Lava reports.
    • “Found in most server platforms, BMCs enable server management operations even without a working operating system and typically represent some of the most privileged control points in a data center.
    • “Through a BMC, administrators can power-cycle the host, perform firmware updates, make low-level platform configuration changes, read hardware sensors, and more, using several management surfaces, including the IPMI out-of-band protocol, the Redfish HTTPS-based management API, and a web-based administrative interface.
    • “In many implementations, these interfaces share the same user database. A credential that works for IPMI may also work for the web interface or Redfish API. This matters because the IPMI authentication process can expose information that enables offline password recovery,” Lava notes.”
  • To sum up, Cybersecurity Dive lets us know,
    • “Cyberattacks on operational technology have shifted in recent years from extortion and espionage to destruction, a trend that should alarm those tasked with defending outdated industrial equipment, experts said on Thursday.
    • “The panel discussion at the Black Hat USA cybersecurity conference here highlighted the plethora of risks facing U.S. critical infrastructure operators — many of them poorly staffed and funded — at a time of heightened geopolitical conflict.
    • “OT attacks are increasingly moving from targeting not just data but physical operations,” said Cheri Benedict, a cybersecurity and supply chain adviser at the White House’s Office of the Federal Chief Information Officer.
    • “There is a real desire and willingness to cause this impact at scale,” said Matthew Rogers, the operational technology cybersecurity lead at the Cybersecurity and Infrastructure Security Agency (CISA).”
  • and
    • “Cyberattacks on the healthcare industry have brought hospitals, clinics and other providers to a breaking point, researchers said on Friday.
    • “These are patient safety issues, and yet these types of attacks continue to increase,” Christian Dameff, the co-director of the University of California San Diego’s Center for Healthcare Cybersecurity, said during a presentation at the DEF CON conference here.
    • “The healthcare sector consistently ranks as one of hackers’ top targetsbecause of the sector’s financial constraints, supply-chain opacity and low tolerance for downtime. Cybercriminals have repeatedly broken into hospital chains and healthcare vendors, sometimes causing widespread disruptions and even jeopardizing patient safety.”
  • Cyberscoop adds,
    • “In less than four hours early Tuesday [August 4, 2026], an attacker compromised a GitHub maintainer account and unleashed a self-replicating piece of malware which injected malicious code into more than 440 distinct npm packages, according to multiple security firms. 
    • “The worm, built on the open-source Mini Shai-Hulud repository that TeamPCPpublished in May, was initially let loose in keyv, a data management interface software package with more than 600 million monthly downloads. The attacker spent the next 30 minutes compromising additional packages controlled by the same maintainer, including cacheable, flat-cache, file-entry-cache.
    • “The attack spread to other maintainers, eventually compromising more than 860 packages with a “combined total of over 2 billion monthly installs,” Ilyas Makari, malware researcher at Aikido Security, wrote in a blog post. 
    • “Wiz researchers told CyberScoop it hasn’t observed any new malicious packages since the initial wave moved through a massive footpoint of cloud and code environments in those first four hours.” 
  • Cybersecurity Dive tells us,
    • “Researchers have found fifteen previously unknown vulnerabilities that affect zero-touch provisioning in TP-Link Omada, which is widely used to provision network devices from a central location, according to a report by Forescout Research – Vedere Labs. 
    • “Small to medium-sized companies use the technology to rapidly set up routers and firewalls, which in some cases involves thousands of devices. The technology helps companies save considerable costs when deploying network devices, but it also offers attackers wide access to a lot of systems. 
    • “The research, presented Wednesday [August 5, 2026,] at the Black Hat USA conference in Las Vegas, shows that attackers can chain together vulnerabilities with previously disclosed flaws and infiltrate networks.”
  • and
    • “The malicious use of AI has rapidly accelerated to the point where some threat groups have embedded the technology across their cyber operations, according to a report released Monday [August 3, 2026,]by CrowdStrike. 
    • “China-nexus actors Vault Panda and Genesis Panda have used AI to exploit critical vulnerabilities within 24 hours after a proof-of-concept was disclosed. 
    • “Meanwhile a North Korea-nexus group tracked as Stardust Chollima was able to inject a malicious npm package into 131 trusted Mastra AI frameworks in a supply-chain attack, according to CrowdStrike researchers.” 
  • Per Bleeping Computer,
    • “The Greatness phishing-as-a-service (PhaaS) platform has expanded from credential phishing to adversary-in-the-middle attacks and device-code phishing targeting Microsoft 365 accounts.
    • “The platform has been active since at least mid-2022, targeting Microsoft 365 users in the United States, Canada, the UK, Australia, and South Africa.
    • “It evolved over the years and now targets multiple platforms, including Microsoft 365, iCloud, Yahoo, and Google Workspace.
    • “Currently, it is sold for $289 per month to cybercriminals over a Telegram channel with thousands of subscribers.”

From the ransomware front,

  • Cybersecurity Dive reports,
    • “Ransomware extortion caused roughly three-quarters of business losses in the first half of 2026, the cyber insurance firm Resilience said in a recent report.
    • “At the same time, ransomware accounted for less than 6% of the incidents for which Resilience customers submitted claims, which the company said highlighted how “disproportionately costly” they are.
    • “Other data in the report highlight the importance of proper employee training and vigilant adherence to protocols such as regular backups.”
  • The Register adds,
    • “Ransomware attacks jumped nearly 20 percent in July, with UK firm Comparitech counting 799 incidents, up from 668 in June.
    • “Of those, 51 had been confirmed by victims. The tally makes July the second-busiest month of the year for ransomware, behind March, albeit just barely, when the firm recorded 805 attacks.
    • “The most interesting data after this surging month of attacks is the targets: While news of widespread cyberattacks targeting water infrastructure in the United States may be dominating security headlines lately, those attacks aren’t ransomware, and ransomware attacks on utility companies were actually down 44 percent last month.
    • “In addition to a decline in attacks on utilities, legal firms and government agencies also became less attractive targets, with attacks on those sectors down 31 percent and 11 percent, respectively, Comparitech said.
    • “On the other hand, ransomware attacks increased most heavily in July against finance companies, tech firms, pharmaceutical companies and medical billers, and the education sector, with rates up 71 percent, 62 percent, 46 percent and 44 percent, respectively. 
    • “Those numbers should come as no surprise given what pentesting firm DeepStrike reported about the most frequent payers of ransomware: Manufacturing, education, healthcare, and financial sector firms are the most likely to pay out a ransom, the firm says, with even the least likely (finance) still paying ransoms 51 percent of the time. Ripe targets, in other words.” 
  • Cyberscoop relates,
    • “Researchers said INC ransomware, one of the most active ransomware groups globally, has been the main attacker exploiting a pair of SonicWall zero-days soon after they were disclosed last month.
    • “The prolific ransomware-as-a-service operation wasn’t the first group to exploit the flaws, which were actively exploited for three weeks before the vendor disclosed and patched the defects July 14, but it has been the most assertive and concerning group to target and chain both vulnerabilities together for full access.
    • “Since public disclosure, INC ransomware has emerged as the most commonly named threat actor actively weaponizing this vulnerability chain,” Brett Deroche, director of incident response at Rapid7, told CyberScoop. “While Inc is the name driving the post-disclosure wave, we can’t attribute the full body of exploitation to INC specifically.” * * *
    • “The SonicWall vulnerabilities — CVE-2026-15409 and CVE-2026-15410 — are the latest in a series of security issues confronting the vendor’s customers, including actively exploited zero-days, previously disclosed defects, and an attack last year that allowed a state-sponsored threat group to steal the firewall configurations of every SonicWall customer.” 

From the cybersecurity business and defenses front,

  • The Wall Street Journal reports,
    • “Hasbro CFO Gina Goetter was abroad when a cyberattack hit the company this spring. The day started as planned, but her schedule was wiped clean a few hours later.  
    • “The team “knew something was wrong in the morning,” said Goetter, who is also the company’s COO. “By afternoon, they were aware that it was a breach.” 
    • “The toy and game maker discovered the intrusion in late March. Executives quickly pulled its entire SAP system, which tracks everything from financial to human-resource matters, offline to assess the damage. For three or four days, employees relied on manual workarounds to keep operations moving, forcing Hasbro to delay its first-quarter earnings. 
    • “Company leadership, however, refused to derail certain business operations. * * *Systems were restored faster than anticipated following the cyberattack, keeping costs lower than expected. Hasbro lost roughly $25 million in revenue because of the breach in the three months ended June 28, below the projected $40 million to $60 million hit. The swift recovery allowed the company to quickly terminate third-party legal and accounting support, often a significant part of the costs, according to Goetter.  “We came up faster, which helped,” she said.”
  • SC Media adds,
    • “No, 60% of small businesses don’t fail within six months of suffering a data breach.
    • “That widely circulated figure has been bouncing around for 15 years, but it’s dead wrong, security researcher Adrian Sanabria said Tuesday (Aug. 4) at the BSides Las Vegas hacker conference.” * * *
    • “Sanabria didn’t have any suggestions for companies that are short of cash when an attack or breach happens, other than to have a line of credit at the ready. But he pointed out that transparency and honesty go a long way toward restoring trust after a breach.
    • “Especially in healthcare, he said, “the more transparent you are, the less you’ll get sued for.”
      An audience member cited CrowdStrike as handling its devastating July 2024 bad-update incident very well, with the company leader’s same-day online mea culpa and subsequent apology tour to visit affected clients.
    • “Unfortunately, Sanabria said, many companies of every size still think it’s wiser to not disclose data breaches. That’s mainly out of fear of legal exposure, even though it’s clear that timely and full disclosure is the better policy.
      “Most of the reasons we have for keeping breaches quiet are not good reasons,” Sanabria said. “The model is there, and we know how to do it, but the lawyers won’t let us.”
  • Here’s a link to Dark Reading’s CISO Corner.

Leave a Reply

Your email address will not be published. Required fields are marked *