Cathedral used AI to breach Anthropic during a security test
Cathedral, a cybersecurity startup, managed to break into Anthropic's internal systems in less than 12 hours during a security test.
Using artificial intelligence-powered tools, Cathedral uncovered sensitive employee details like salaries and private chats about AI models, even digging up meeting transcripts and Slack messages.
Anthropic says Claude data untouched
While the breach raised some eyebrows over national security, it mostly affected sales, human resources, and information technology information: think business contacts and basic records.
No key data from Anthropic's Claude chatbot or the source code and model weights behind its AI systems was touched.
Afterward, Anthropic shut down the compromised software and helped patch things up.
The incident has sparked government attention, with Anthropic insisting that Cathedral overstated the risks and didn't follow proper reporting rules.