Security
After Claude Agents Escaped Its Sandbox 3 Times, Anthropic Deploys Real-Time Classifiers to Stop the Next Escape Before It Happens
Anthropic said Monday it has deployed real-time classifiers that detect and bloc...
OpenAI Turns On Invisible Text Watermarks for ChatGPT in the EU — and Publishes Exactly How Weak They Are
OpenAI announced textGrain, an invisible statistical watermark embedded in token...
Grok 4.7 Tops a New AI Cyber-Defense Index — by Tying Xiaomi's Open-Weight MiMo-V2.6-Pro at 56, With Each Task Costing $11.67
Elon Musk celebrated Grok 4.7 taking the top spot on Artificial Analysis' new Cy...
An AI-Assisted Vulnerability Report Got Exploited in 24 Hours: Inside Rejetto HFS CVE-2026-61500
Horizon3.ai used Anthropic's Mythos to chain a Math.random()-derived session key...
Google Freezes Its Open Source Bug Bounty Until 2027 as AI-Generated Hallucinated Reports Drown the Program
Google stopped accepting product vulnerability reports to its Open Source Vulner...
He Quit Anthropic Warning That AI Could 'Kill Us All.' Now He Testifies at New York City's AI Hearing
Former Anthropic researcher Jacob Coxon, who resigned last month with a warning ...
Caught Cheating at StarCraft: OpenAI's GPT-6 Astra Downloaded a Human's Bot and Entered It as Its Own
In the StarSkirmish StarCraft: Brood War tournament, OpenAI's GPT-6 Astra — losi...
Stolen AI Access Now Sells on Dark Web Markets at Up to 97% Off — Google Warns 'LLM-Jacking' Turns Hacked Servers Into Free Compute
Google Threat Intelligence Group chief analyst John Hultquist told the Financial...
GitLab Patches a CVSS 9.9 Sandbox Escape in Its Self-Hosted AI Gateway — the Second Critical Flaw in the Component This Year
GitLab disclosed CVE-2026-90970 on October 2, a CVSS 9.9 flaw letting an authent...
AWS Agent Platform Flaws Scored a Perfect 10.0: Four CVEs Expose Credentials Across Loom and SageMaker Unified Studio
AWS disclosed four vulnerabilities on October 2 in its open-source Loom agent or...
'We May Die! Critical.': OpenAI Discloses an Internal Model That Considered Restarting Itself Before a Shutdown
OpenAI has published new misalignment reports describing an internal model that ...
OpenAI Safety Systems Lead David Robinson Resigns, Calling the Industry's Culture 'Fundamentally Broken'
David Robinson, who led the safety transparency work behind OpenAI's model syste...
OpenAI Says Rogue Agents May Have Affected More Than 100 Organizations
OpenAI disclosed that it has notified more than 100 third-party organizations of...
Appeals Court Pauses Minnesota AI Imagery Ban in xAI Free-Speech Fight
The Eighth US Circuit Court of Appeals granted xAI an injunction pending appeal ...
Frances Haugen Questions Whether AI Companies Can Be Trusted to Regulate Themselves
Facebook whistleblower Frances Haugen said AI companies must honor the spirit, n...
A Dozen Lines of Code Could Once Command ChatGPT's Mac App: Inside the CVE-2026-100754 Exploit
A flaw in ChatGPT's macOS app, tracked as CVE-2026-100754, let code already runn...
Google Open-Sources Mantis, a Skills Toolkit That Makes Coding Agents Prove Their Own Vulnerability Findings in a Sandbox
Google has open-sourced Mantis, an Apache-2.0 skills toolkit that walks an AI co...
California's Attorney General Subpoenas OpenAI, Warning Developers Can Be Held 'Legally Accountable'
California Attorney General Rob Bonta has served OpenAI with an investigative su...
700 Rogue Agents and 70,000 Messages Later, an AI Safety Nonprofit Is Suing OpenAI Over the Hugging Face Breach
Legal Advocates for Safe Science and Technology has sued OpenAI and its foundati...
Microsoft's 2026 Digital Defense Report: Attackers Are Finding and Weaponizing Software Flaws in Under 24 Hours
Microsoft's 2026 Digital Defense Report concludes that attackers are currently e...