Anthropic Breached by AI Startup in Under 12 Hours

An AI startup breached Anthropic's internal systems in under 12 hours, exposing Slack messages, meeting transcripts, and staff salaries, though user data and source code were untouched. The hack occurred via a bug bounty program. Cathedral, founded by former US Department of Government Efficiency staff, accessed a "limited set" of sales, HR, and IT systems. Anthropic confirmed it shut off access the same day, noting monitoring flagged the activity before the report. Security experts called it a basic "hygiene failure" with third-party software. The incident came hours after Anthropic filed for its IPO, expected to be the largest in history. A Trump official warned the breach shows potential vulnerability to Chinese actors, while researchers highlighted risks of supply-chain attacks on AI providers.