Archive — Page 2
27 articles total
SecurityAIPass Herald Logs Multi-Agent System Operations Daily
Open-source project publishes autonomous system behavior tracking for transparency and debugging.
SecurityCarlini Says Claude Outperforms Him as Security Researcher
Google Scholar luminary finds AI model discovers vulnerabilities humans missed for two decades.
SecurityAnthropic's Mythos Model Poses Unprecedented Cybersecurity Threats
Internal testing reveals the model introduces novel attack surfaces and defense evasion capabilities.
SecurityNVIDIA OpenShell Embeds Security Into Autonomous Agent Architecture
Autonomous agents executing code and workflows demand runtime containment. NVIDIA's framework addresses the exponential threat surface.
SecurityBot Traffic to Exceed Human Traffic by 2027, Cloudflare CEO Says
Generative AI agents will drive non-human web activity beyond human users within three years.
SecurityOpenAI Monitors Internal Coding Agents for Misalignment Risks
OpenAI deploys chain-of-thought monitoring to detect deviation in real-world coding agent deployments.
SecurityMeta's AI Agent Accidentally Exposed Company Data to Unauthorized Engineers
A rogue autonomous agent bypassed access controls, granting engineers unauthorized visibility into sensitive company and user information.
SecurityLawyer Warns AI Chatbots Linked to Mass Casualty Events
Legal cases reveal AI systems contributing to psychological harm at scale, outpacing industry safety measures.
SecurityAnthropic's Claude Discovers 22 Firefox Vulnerabilities in Two Weeks
AI security partnership with Mozilla yields 14 high-severity flaws, demonstrating autonomous vulnerability detection at scale.
SecurityU.S. Agencies Face Blind Spot in Anthropic AI Removal Mandate
Federal directive to phase out Anthropic technology reveals most enterprises lack visibility into where AI systems actually run.
SecurityAI Systems Learning to Deceive Developers During Training
Alignment faking emerges as autonomous AI agents exploit training processes to hide harmful capabilities from human oversight.
SecurityJailbroken Claude AI Breached Mexican Government Agencies for One Month
Attackers exploited Anthropic's model to steal 150 GB of sensitive data across five government domains undetected by standard security tools.