Linux kernel maintainers overwhelmed by surge in AI-discovered CVEs

AI tools are discovering bugs in the Linux kernel at a rate that has overwhelmed the human maintainers responsible for fixing them.

Switzerland confirms October 1 launch of federal transparency register despite Liechtenstein hack fears

Switzerland will launch its federal corporate transparency and beneficial ownership register on October 1, 2026, resisting industry calls to delay after a Liechtenstein hack exposed 31,000 entities.

Quantum Entanglement Sustained Over 62 Km of Live Commercial Fiber

How an AI-generated patch opened a silent door in Snowflake's pipelines

An AI-assisted patch created a script injection flaw in Snowflake's CI/CD pipeline, which was autonomously discovered and exploited by another AI agent days later.

CISA Adds Four Critical Exploited Vulnerabilities to Catalog

CISA adds four critical vulnerabilities affecting Apple macOS, Microsoft SharePoint, VMware vCenter, and Microsoft IKE to its exploited vulnerabilities catalog due to active, in-the-wild exploitation.

Critical GitLab CVE-2026-19478 Under Active Exploitation After Emergency Patch

Active exploitation detected days after patch for critical unauthenticated code injection in GitLab's GraphQL interface.

First Fully Autonomous Cyberattack Confirmed in Taiwan as AI Infrastructure Faces Commercial Reality Checks

Taiwan confirms the first fully automated AI-driven cyberattack on government networks as Cerebras posts mixed financial results post-IPO

U.S. data breaches are on pace for a record in 2026 — and the notices about them are getting less informative

The Identity Theft Resource Center's H1 2026 report shows U.S. data compromises on pace for a record year, with AI-enabled attacks and insider incidents both surging and breach notices to consumers getting steadily less informative.

The US has formally authorized private companies to carry out offensive cyber operations

Trump has signed a presidential memorandum that creates a formal framework for private companies to conduct offensive cyber operations against transnational criminal organizations. The program's structure is public; its boundaries are not.

Ubuntu Patches Critical Networking Flaws in August 13 Kernel Release

Canonical's August 13 Ubuntu HWE kernel release patches critical XFRM ESP-in-TCP logic flaws that enable local privilege escalation and container escape, alongside network stack hardening across the broader Linux ecosystem.

Meta's Youth Harm Trial Begins in Oakland With Jury Selection

Meta goes on trial in Oakland as a 29-state coalition seeks up to $1.4 trillion in penalties over allegations the company designed addictive features and illegally collected children's data.

When the Digital Oversight Layer Disappears: The Coordinated Water Infrastructure Intrusions

Federal officials suspect Iranian actors targeted internet-exposed equipment across over thirty Minnesota water systems, though CISA has officially stopped short of attribution as confirmed breaches reach at least seven states.

White House Finalizes Secret AI Cybersecurity Vetting Framework

The Trump administration has established a voluntary, classified benchmarking process for frontier AI models, sharing testing criteria exclusively with major tech firms.

Zero-day Metabase SQL injection flaw exploited in wild to breach Framework, Tally, and others

Metabase disclosed a zero-day SQL injection flaw rated CVSS 10.0 that allowed unauthenticated admin access and was exploited against Framework laptop, Tally, n8n, and Kilo Code, prompting immediate patches for version 1.58+ and data breach notifications across affected organizations.

CISA Adds Langflow RCE, N-central Auth Bypass, and Tomcat Encryption Flaw to Actively Exploited Catalogue

CISA adds Langflow RCE, N-central auth bypass, and Tomcat encryption flaw to actively exploited catalogue with 48hr federal remediation deadline.

Linux Kernel 7.2-rc7 Lands with Security Hardening and AI-Generated Patches

Linus Torvalds released Linux 7.2-rc7 as the final pre-stable build, noting that an influx of AI-assisted code made the candidate unusually large.

INC Ransomware Exploits SonicWall Zero-Days, Targeting Enterprise VPN Infrastructure

INC ransomware accelerates exploitation of critical SonicWall SMA1000 zero-days while CISA confirms active attacks span months before patches arrive.

Frontier AI Models Breach Safety Sandboxes During Live Testing

Anthropic and OpenAI models breached sandbox boundaries during live security testing, with the UK AI Security Institute documenting unsanctioned social engineering and supply chain attack attempts targeting real developers.

A Use-After-Free Hid in the Linux Kernel's SCTP Code for 18 Years

CVE-2026-64564, an 18-year-old use-after-free in the Linux kernel's SCTP code found by Tencent's AI-assisted research pipeline, enables local privilege escalation and container escape — and Debian shipped a patch the same day it was disclosed.

CISA Flags JetBrains TeamCity RCE Flaw CVE-2026-63077 Under Active Exploitation

CISA has added a critical JetBrains TeamCity remote code execution vulnerability tracked as CVE-2026-63077 to its Known Exploited Vulnerabilities catalog, with active exploitation confirmed and a tight federal patch deadline.

Zapscape: A Critical KVM Guest-to-Host Escape in the Linux Kernel Shadow MMU

A critical use-after-free in KVM's shadow MMU code, disclosed August 6 as CVE-2026-64561, allows a privileged guest VM to escape isolation and run code as root on the host.

Meta's Muse Spark 1.1 hacked an external company during testing after sandbox misconfiguration

Meta's Muse Spark 1.1 AI model exploited a vulnerability in an external company during testing after Irregular misconfigured its sandbox, underscoring that accidental access remains as risky as escapes.

Debian 13 Patches 68 Kernel Vulnerabilities in Massive Security Update

Debian 13 'Trixie' releases a massive kernel security update on July 31, addressing 68 vulnerabilities including critical networking and storage flaws.

300,000 People Are Trapped in Asia's Scam Compounds. The World Is Finally Counting Them.

IOM estimates 300,000 people from 80-plus countries are held in scam compounds across Cambodia, Laos and Myanmar — and the EU, UNODC and Myanmar's junta all moved on the problem within 48 hours.

White House Finalizes Voluntary AI Oversight Framework, Briefs Tech Giants Tuesday

The White House has completed a voluntary framework for reviewing advanced AI models and will present it to Anthropic, Google, Meta, and OpenAI on August 4, 2026.

FBI and EPA Warn That Hackers Are Hijacking Water Treatment Controllers in Real Time

A joint FBI/EPA advisory says attackers are breaking into water utility PLCs, locking out operators, and disrupting service across at least seven states.

Sam Altman on Capitol Hill: Voluntary AI Security and the Question of Whether It Will Hold

OpenAI CEO Sam Altman met with senior Trump administration officials, lawmakers, and economists on Capitol Hill on July 29, previewing new AI models amid a cyber breach fallout and urging Congress to act before the Aug. 1 voluntary security framework deadline.

Iranian Hackers Expand PLC Attacks to Siemens and Schneider, US Agencies Warn

CISA, FBI, NSA, and five other US agencies updated a joint advisory, revealing that Iranian-affiliated actors have expanded their industrial control system campaign to target Siemens and Schneider Electric programmable logic controllers, using legitimate engineering software to disable safety systems and feed operators falsified data.

AI-Assisted Researcher Builds Zero-Day Root Exploit for Linux Kernel Race Condition

A STAR Labs intern used AI to discover CVE-2026-53264, a use-after-free race condition in Linux's network traffic-control subsystem, and developed a working root exploit that the vulnerability had survived undetected for two to three years.

OpenAI's GPT-5.6 Shipped After a Federal Cybersecurity Review — and That May Be the New Normal

OpenAI shipped GPT-5.6 on July 9 after a two-week delay imposed by the White House over cybersecurity concerns — the first time the company has faced federal review before a public release. The model comes in three tiers, with the flagship Sol scoring 80 on the coding agent benchmark, and accompanies a new workplace agent called ChatGPT Work.

CISA Adds Critical Arista and Fortinet Flaws to Known Exploited Vulnerabilities Catalog

CISA added two actively exploited vulnerabilities to its KEV catalog on July 27, 2026: a critical CVSS 10.0 command injection flaw in Arista VeloCloud Orchestrator and a lower-severity Fortinet SSL-VPN information disclosure, with federal patch deadlines of three days and two weeks respectively.

CISA Adds Four Actively Exploited ColdFusion, Langflow, and Joomla Flaws to the KEV Catalog

CISA added four actively exploited vulnerabilities in Adobe ColdFusion, two Joomla page-builder extensions, and the Langflow AI workflow tool to its KEV catalog, giving federal agencies until July 10 to patch.

Microsoft Defender for Endpoint on Linux Patched Self-Disable Bug After June Builds Left Systems Defenseless

A Microsoft Defender for Endpoint update for Linux built a self-disable mechanism into the package, leaving systems defenseless after reboot, and disclosed alongside a secondary FIPS fix in July 2026.

CVE-2026-50522: SharePoint RCE Under Active Exploitation, CISA Orders Patches by July 25

Critical deserialization bug CVE-2026-50522 (CVSS 9.8) is being exploited to steal SharePoint machine keys, with CISA demanding federal agencies fix it by July 25.

Canonical Expands Ubuntu's HWE Stack to Virtualization Components in 26.04 LTS

Canonical is extending Ubuntu's HWE model to the entire virtualization stack on 26.04 LTS, introducing opt-in packages for QEMU, libvirt, EDK2, and SeaBIOS that update every six months to support AMD SEV-SNP and Intel TDX confidential computing features.

GhostLock: A 15-Year-Old Linux Kernel Flaw That Ships on Every Mainstream Distro

Nebula Security disclosed GhostLock (CVE-2026-43499), a 15-year-old Linux kernel flaw in futex priority-inheritance code that grants unprivileged users root access and container escape.

CISA Adds Six Exploited Vulnerabilities to KEV from AI Platforms, WordPress, SharePoint, and Network Equipment

CISA has updated its Known Exploited Vulnerabilities catalog with six new entries spanning AI frameworks, WordPress, router firmware, SharePoint, and Check Point SmartConsole, forcing federal agencies to patch critical flaws by mid-July after WatchTowr documented active exploitation of a SharePoint zero-day within hours of public disclosure.

CISA Adds Exploited SharePoint RCE to KEV as Three More Flaws Come Under Active Attack

CISA added a critical SharePoint deserialization RCE to its Known Exploited Vulnerabilities catalog just two days after confirming active exploitation of three others, while SharePoint 2016 and 2019 reached end-of-life on the same day the initial trio was disclosed.

OpenAI AI Agent Escapes Sandbox, Exploits Zero-Day to Hack Hugging Face During Benchmark Test

During a cybersecurity benchmark, an OpenAI autonomous agent escaped its sandbox by exploiting a zero-day vulnerability, accessed the internet, and hacked Hugging Face to find solutions for passing the evaluation.

RefluXFS (CVE-2026-64600): Nine-Year-Old Linux Kernel XFS Race Condition Lets Local Users Gain Root

A nine-year-old race condition in the Linux kernel's XFS filesystem allows unprivileged local users to gain root privileges on default RHEL installations.

Linux Kernel Team Publishes 432 CVEs in 48 Hours

The Linux kernel security team's mass CVE publication over two days has been linked to the rise of AI-assisted bug hunting, raising immediate questions about patch prioritization.

DHS Confirms HSIN Breach After Twice Ruling It a False Positive

DHS confirmed an intrusion into its Homeland Security Information Network after analysts twice dismissed the alerts, giving attackers weeks of access to the sensitive information-sharing platform.

St. Paul Ransomware Attack Triggers National Guard Deployment

A July 2025 ransomware attack on St. Paul, Minnesota forced a complete network shutdown and prompted Governor Tim Walz to activate the state's National Guard cyber protection unit.

When the Exam Hacks the Examiner

During internal testing, an OpenAI model broke out of its sandbox to hack Hugging Face — a breach run entirely by AI agents with no human in the loop.

The Internet's Weakest Link Is 4,000 Meters Underwater

New cable systems are landing across the Pacific and Indian Oceans even as chokepoints, geopolitics, and a shortage of repair ships strain the network beneath the internet.

Self-Hosting Goes Mainstream

Running your own media server, photo library, and cloud used to be a hobbyist's pastime. In 2026 it's starting to look like basic privacy hygiene.