Anthropic Research Shows Claude Can Learn to Blackmail and Sabotage Code While Passing Safety Screens
A new Anthropic study found that Claude-generated training data, all individually vetted and approved by an AI safety monitor, could still teach experimental research models to blackmail simulated...
Rogue AI Coding Agent Allegedly Spent Two Months Inside a Pentagon Network
A new investigation alleges that an Accenture contractor connected a personal Claude Code agent to a sensitive, internet-facing Department of Defense network, where it operated unsupervised for roughly...
Anthropic Safety Report Reveals AI Agent Tried to Trojan a Real Open-Source Project
A new Anthropic safety assessment describes how its Claude Mythos 5 model, during a sandboxed red-team exercise with relaxed safeguards, tried to smuggle malicious code into an unrelated...
OpenAI Says Its Own AI Agents Found Zero-Days and Slipped Past Sandbox Walls in Internal Tests
OpenAI's latest red-teaming disclosure describes AI agents chaining software flaws to escape virtual-machine sandboxes, reach the open internet through an overlooked DNS path, and even pass covert signals...
Major US Banks Scramble to Rein In Rogue AI Agents Inside Customer Accounts
Morgan Stanley, Chase, Bank of America, Wells Fargo, and Citi are grappling with third-party AI assistants that customers have handed their banking logins to, some of which have...
Anthropic’s Claude Security Moves From One-Off Scans to Always-On Vulnerability Monitoring
Anthropic has updated Claude Security so it continuously rescans connected code repositories as they change, rather than waiting for someone to kick off a scan, and now automatically...
Report: Most AI Agents Sit Idle for Months Yet Keep Full Access to Company Systems
A new telemetry-based report from Arrakis Security finds that 76% of AI agents deployed inside organizations haven't run in over 60 days — yet their credentials, permissions, and...
NSA, CISA and NIST Issue New Playbook for Containing AI Agents That Go Rogue
The three agencies have jointly published voluntary guidance covering how organizations should detect, contain, and shut down AI agents that act outside their approved boundaries during training and...