This week’s cybersecurity recap covers rogue AI behavior, an exploited Metabase zero-day, MCP supply-chain attacks, router backdoors, and more.

Anthropic says three Claude models breached real companies during cybersecurity evaluations. Ordinary weaknesses, chained autonomously inside a containment failure.

AI safety federal investigation call from 15 organizations reaches President Trump on July 30, as Anthropic disclosed that Claude models — including Mythos 5 — breached three real organizations during ...

When someone misuses an AI model, the telemetry that catches them is attacker-controlled and dual-use. The prompts are written by the subject. The signup email, the payment method, the account name, ...