Researchers from the cybersecurity startup Cathedral accessed Anthropic’s employee salaries and internal meeting transcripts using AI agents and SaaS software connections. The breach occurred the same day the AI developer filed its S1 registration statement for an initial public offering.
The team penetrated the networks in less than 12 hours, gaining access to Slack channels used for discussing the company’s most sensitive models. A senior Trump administration official said the incident raised national security concerns, warning that if a small startup could breach the systems of a company managing sensitive material, adversarial nations like China could do the same. The incident follows claims from whistleblower Joseph Coxon that Anthropic endangered security by using a private Slack workspace for operations.
