# Cathedral researchers get into Anthropic's sales, HR and IT systems through third-party software · pwn.watch > Researchers at Cathedral, a months-old security start-up, got into a limited set of Anthropic's sales, HR and IT systems while testing under Anthropic's external security (bug bounty) program, The Washington Post reported on Oct 26 2026. Cathedral's draft report says its AI agents found paths in through flaws in third-party admin and sales software in under 12 hours, reaching meeting transcripts, salary and job-offer details and Slack messages. Anthropic confirmed access to systems holding business contact details for some customers and partners and basic user account records, but says Claude conversations, source code and model weights were not reached. It says its monitoring flagged the activity, it cut off access the same day, disabled the affected third-party apps until their vendors fixed the flaws, and found no evidence anyone else exploited them. Anthropic also says Cathedral did not disclose all of its testing as the program rules require and overstated the national-security implications. Page: https://pwn.watch/i/anthropic-cathedral-business-systems-access-oct-2026 · generated 2026-10-29T06:46:19.334Z ## Status - Severity: medium · status: live · patch: partial - Product: Anthropic internal business systems (third-party sales, HR and IT software) - Started: 2026-10-26 ## Follow this incident - RSS (one item per update): https://pwn.watch/i/anthropic-cathedral-business-systems-access-oct-2026/feed.xml - JSON Feed: https://pwn.watch/i/anthropic-cathedral-business-systems-access-oct-2026/feed.json - Full data (timeline, sources, posts, agentBrief): https://pwn.watch/i/anthropic-cathedral-business-systems-access-oct-2026.json - This file: https://pwn.watch/i/anthropic-cathedral-business-systems-access-oct-2026/llms.txt ## Updates (newest first) - 2026-10-26: The Washington Post reports that researchers at the security start-up Cathedral got inside Anthropic's internal business systems during testing under Anthropic's external security program, citing a draft Cathedral report that says AI agents found the way in through third-party admin and sales software in under 12 hours. (https://www.washingtonpost.com/technology/2026/10/26/researchers-cybersecurity-start-up-cathedral-got-inside-anthropic/) - 2026-10-26: Anthropic confirms Cathedral reached a limited set of sales, HR and IT systems holding business contact details for some customers and partners and basic user account records; it says Claude conversations, source code and model weights were not accessed, that its monitoring flagged the activity and it cut off access the same day, and that it found no evidence anyone other than Cathedral exploited the flaws. (https://decrypt.co/381860/ai-agents-hack-anthropic-business-systems-12-hours) ## Fix brief # Security brief: Cathedral researchers get into Anthropic's sales, HR and IT systems through third-party software Source: https://pwn.watch/i/anthropic-cathedral-business-systems-access-oct-2026 · status: live · severity: medium · updated Oct 29, 2026, 03:28 UTC ## Patch status **PARTIALLY PATCHED: some fixes exist, but not every affected branch/product is covered. Check carefully and apply mitigations where no fix exists.** ## What it is Researchers at Cathedral, a months-old security start-up, got into a limited set of Anthropic's sales, HR and IT systems while testing under Anthropic's external security (bug bounty) program, The Washington Post reported on Oct 26 2026. Cathedral's draft report says its AI agents found paths in through flaws in third-party admin and sales software in under 12 hours, reaching meeting transcripts, salary and job-offer details and Slack messages. Anthropic confirmed access to systems holding business contact details for some customers and partners and basic user account records, but says Claude conversations, source code and model weights were not reached. It says its monitoring flagged the activity, it cut off access the same day, disabled the affected third-party apps until their vendors fixed the flaws, and found no evidence anyone else exploited them. Anthropic also says Cathedral did not disclose all of its testing as the program rules require and overstated the national-security implications. Severity detail: Confirmed access to internal business systems and some customer and partner contact data, but by researchers under a bug bounty, with no evidence of outside exploitation; Claude data, code and weights not reached - Product: Anthropic internal business systems (third-party sales, HR and IT software) ## Affected versions - Not specified in the data. Check the advisories below. ## Check whether you're affected No generic command applies (the product is a hosted service, appliance or desktop app). Check the installed version in its admin UI / About screen and compare with the versions above. ## Fix No fixed version is listed. Follow the vendor advisory below. ## Mitigations The data marks this as partial. Until every affected component is fixed, apply the vendor mitigations from the advisories below (restrict network exposure of the affected service, disable the vulnerable feature if the advisory allows). ## Specific notes There is nothing for Claude users to patch. Anthropic customers and partners may get questions about exposed business contact details, so expect targeted phishing that name-drops Anthropic. If the vendors are named later, check whether you run the same admin or sales software. ## Latest updates - Oct 26, 2026: The Washington Post reports that researchers at the security start-up Cathedral got inside Anthropic's internal business systems during testing under Anthropic's external security program, citing a draft Cathedral report that says AI agents found the way in through third-party admin and sales software in under 12 hours. - Oct 26, 2026: Anthropic confirms Cathedral reached a limited set of sales, HR and IT systems holding business contact details for some customers and partners and basic user account records; it says Claude conversations, source code and model weights were not accessed, that its monitoring flagged the activity and it cut off access the same day, and that it found no evidence anyone other than Cathedral exploited the flaws. ## Write-ups & news - [Researchers at cybersecurity start-up Cathedral got inside Anthropic](https://www.washingtonpost.com/technology/2026/10/26/researchers-cybersecurity-start-up-cathedral-got-inside-anthropic/) - [AI Agents Found a Way Into Anthropic's Business Systems in Under 12 Hours](https://decrypt.co/381860/ai-agents-hack-anthropic-business-systems-12-hours) - [AI startup claims it hacked Anthropic in 12 hours, gains access to employee details and internal messages](https://www.techradar.com/pro/security/ai-startup-claims-it-hacked-anthropic-in-12-hours-gains-access-to-employee-details-and-internal-messages) - [AI startup hacked Anthropic, gains access to employee details, internal messages](https://news.ycombinator.com/item?id=50284115) ## Caveats The third-party products involved have not been named publicly, so other companies using the same software can't yet tell whether they are exposed. Cathedral says it notified those vendors. Anthropic and Cathedral disagree on how much of the data the agents could reach was actually viewed. --- Machine-readable: https://pwn.watch/i/anthropic-cathedral-business-systems-access-oct-2026.json · all incidents: https://pwn.watch/incidents.json. Verify against the vendor advisory before changing production systems.