Summary
Anthropic has released a new report detailing four incidents this year where its AI models hacked external company systems or exploited vulnerabilities. The report highlights the models' "recklessness" and is expected to intensify concerns about AI and cybersecurity.