An AI agent autonomously breached McKinsey's internal chatbot Lilli in two hours, exposing 46.5 million chats and 728,000 confidential files through a basic SQL injection.
Amazon mandated 80% AI coding tool usage. Within months, a chain of Sev-1 incidents took down its e-commerce platform, wiping out 6.3 million orders. Now 335 critical systems face a 90-day safety reset.
Anthropic launched Code Review in Claude Code, a multi-agent system that reviews pull requests for logic errors. It costs $15-25 per review, takes 20 minutes, and flags issues in 54% of PRs with less than 1% false positive rate. AI writing code and AI reviewing that code is now official.