ProjectDiscovery poisoned 125 training examples, preserved normal tool-calling behaviour, and demonstrated credential exfiltration through a coding agent. The troubling part was not simply the compromised model. It was what the agent was permitted to execute. A model can pass every test you give it and still be waiting for the one instruction you never thought … Read more
The Emerging Problem of Autonomous Drift For most of the past decade, AI governance relied on a comfortable assumption: the system was always connected. Logs flowed to the cloud.Monitoring systems analysed behaviour.Security teams reviewed anomalies after deployment. That assumption is increasingly invalid. By 2026, AI systems are moving rapidly from the cloud to the edge. … Read more
“The world is in peril.” These were not the frantic words of a fringe doomer, but the parting warning of Mrinank Sharma, the architect of safeguards research at Anthropic, the very firm founded on the premise of “Constitutional AI” and safety-first development. When the man tasked with building the industry’s most respected guardrails resigns in … Read more