If you want to see how AI is changing cybersecurity, look at what happened when a startup pointed its AI agents at one of the world’s leading AI labs.
Researchers at Cathedral said their AI agents found paths into Anthropic’s business systems in less than 12 hours, according to a Washington Post report published on Monday and summarised by Decrypt.
The weak spot wasn’t Claude
The way in wasn’t Anthropic’s AI assistant at all. It was third-party administrative and sales software, the kind of tools companies use to track customers and manage staff. Many other companies use the same software, and the vendors haven’t been named.
Cathedral was founded months ago by four former staffers of the Department of Government Efficiency (DOGE) and raised $160 million at a $1.4 billion valuation in July. It tested Anthropic through a programme that works like a bug bounty. Its agents, which carry out multi-step tasks without a person approving each move, found routes to meeting transcripts, salary and job-offer details, and Slack messages about Anthropic’s advanced models, according to a draft of its report.
Two sides of the story
Anthropic confirmed the researchers reached “a limited set” of sales, HR and IT systems holding business contact details and basic user account records. “Our monitoring flagged some of Cathedral’s activity before they reported it to us,” the company told the Post. “We shut off their access the same day.”
Anthropic says customer conversations with Claude, source code and model weights were not accessed, and it has no evidence anyone other than Cathedral exploited the flaws. It also said Cathedral overstated the national security implications and didn’t fully follow the programme’s rules because it didn’t disclose all its activity. Cathedral’s draft report says “data accesses were kept to the absolute minimum.”
Anthropic disabled the affected software until the vendors confirmed fixes.
A senior Trump administration official, speaking anonymously, said the report had raised national security concerns in government: “If a bunch of kids who have just started a company can do this, then potentially, the Chinese aren’t far behind.”
Why it matters
The lesson for every business is that your security is only as strong as the SaaS tools around your core systems: CRMs, HR platforms, admin panels. AI agents can now probe those connections faster than any human team. Review which third-party apps have access to your data, remove the ones you don’t use, enforce strong authentication and keep an eye on unusual activity. Monitoring is what let Anthropic spot and stop this quickly.
Sources: Decrypt, The Washington Post. Written by the xRapid Media LLP team.