# Anthropic's New AI Found a 27-Year Bug. Then It Wrote a Working Exploit.

- slug: anthropics-new-ai-found-a-27-year-bug-then-it-wrote-a-working-exploit
- date: 2026-04-07
- category: Artificial Intelligence

Anthropic's new AI model found a vulnerability that sat undiscovered in OpenBSD for 27 years. Then it wrote a working browser exploit with no security training. The company is now racing to control what it released.

---