Cathedral claims AI agents breached Anthropic systems under 12 hours
Cathedral, a young startup with ties to Elon Musk's former team, claims it managed to break into Anthropic's internal systems in less than 12 hours using AI agents.
They got into business software and accessed things like meeting transcripts, details about salaries, job offers, and Slack messages.
Anthropic says only some sales, HR, and IT info was touched, and their monitoring flagged some of Cathedral's activity before they reported it to us, and we shut off their access the same day.
Cathedral $1.4B startup alarms security experts
Cathedral is just a few months old but already valued at $1.4 billion and working with US federal agencies.
The fact that such a new company claims to have pulled off this breach has experts worried about how secure AI-powered systems really are:
an unnamed Trump official warns that if startups can do this, the Chinese aren't far behind.