Anthropic is expanding a program that allows vetted cybersecurity professionals to test its most powerful AI models with fewer safeguards, after its Project Glasswing initiative helped uncover more than 100,000 software vulnerabilities this year.
Read full article