Anthropic's Claude writes and proves code safe automatically

Anthropic had Claude generate roughly 180 million lines of proof code showing that existing code in 123 open-source libraries is correct, catching bugs that years of testing missed. The goal is to make AI formal verification as routine as standard testing.

• Experts question Anthropic report on AI reshaping work2026-10-23
• Google Cloud launches Anthropic-managed Claude option2026-10-22
• Claude fails to create scorpion antivenom in experiment2026-10-22
• HHS Plans Sole-Source Anthropic Claude Contract2026-10-21
• Ex-Harvard AI adviser joins Anthropic oversight trust2026-10-21