AI agents from startup Cathedral found paths into Anthropic’s internal systems in less than 12 hours by exploiting weaknesses in third-party administrative and sales software. The researchers accessed meeting transcripts, salary and job-offer information, Slack messages, and a limited set of sales, human resources, and information technology systems. Anthropic said customer conversations with Claude, source code, and model weights were not accessed, and it found no evidence that anyone other than Cathedral exploited the flaws. Cathedral, founded by four former DOGE team members and valued at $1.4 billion, disputed Anthropic’s national security concerns and said it had kept data access to the absolute minimum.
Source: Read the original article
