A bug-bounty probe by AI startup Cathedral reached a limited set of Anthropic’s sales, HR and IT systems, The Independent reports. Anthropic said the incident did not expose Claude users’ conversations or the model’s source code, and that it shut off Cathedral’s access the same day.
Anthropic Watch analysis
What happened
The Independent reports that Cathedral, a company founded by former US Department of Government Efficiency staff, accessed Anthropic’s internal network while looking for security flaws through the company’s bug-bounty programme. The report says Slack messages, meeting transcripts, staff salaries and job offers were exposed. It attributes the original reporting to The Washington Post and quotes Anthropic confirming access to a “limited set” of systems.
Anthropic said its monitoring had flagged some of Cathedral’s activity before the startup reported it. The Independent also reports that Anthropic shut off access on the day it was reported. Its account says Claude conversations and model source code were not accessed.
Why it matters
The reported limits matter: this was not an exposure of Claude conversations or model code. But access to internal business systems is still a serious security incident, particularly for a company whose services are part of other organisations’ workflows. The details also point to a familiar weak spot: security can depend on the tools and systems surrounding a model, not just the model itself.
Our read
This is a substantial, attributed security story, not proof that Cathedral carried out a sophisticated AI attack. The Independent describes a bug-bounty probe, and quotes a security researcher who characterised the issue as a basic failure involving third-party software. That assessment is the researcher’s, not a settled finding. The more useful distinction for readers is what the report says was reached, and what it says was not.
What to watch
- Whether Anthropic or Cathedral publishes a fuller account of the affected systems and the third-party software involved.
- Whether Anthropic discloses any further impact on staff or business operations.
- Whether the incident prompts changes to how AI companies manage access for security researchers.
Discussion spark: When a bug-bounty probe reaches internal systems, how much detail should a company disclose about the exposure, even when customer conversations and model code were not accessed?
Sources and evidence
- Source update (27 October 2026, 14:09 UTC)
Anthropic Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by Anthropic.