X discussion: former OpenAI researcher Will Depue says Anthropic is "massively underplaying" the Cathedral security test, drawing about 2,000 likes and 180K views

What happened

At 00:18 UTC on Oct 28, Will Depue posted on X: "anthropic is massively underplaying the recent security breach. i'm hearing cathedral achieved full administrator access of the slack workspace (see every message, log in as arbitrary users) where literally every single research secret is contained." His X bio says he previously worked at OpenAI on Sora 1 and 2 and on o3/4o post-training. At the time of writing the post had about 2,000 likes, 113 reposts, 621 bookmarks and roughly 184K views, making it one of the most-engaged discussions of the incident over the past half day.

To be clear, the full-admin-access detail is something he says he is "hearing"; neither Anthropic nor Cathedral has publicly confirmed it.

Background

According to a Washington Post report on Oct 26, Cathedral, a security startup founded by four former DOGE staffers, tested Anthropic through its bug-bounty style external security program. Its draft report says its AI agents found paths, through vulnerabilities in third-party administrative and sales software, to meeting transcripts, compensation and offer details, and Slack messages about advanced models in under 12 hours.

Anthropic told reporters the researchers reached only a "limited set" of sales, HR and IT systems, including some customer and partner business contacts and basic user account records; Claude user conversations, source code and model weights were not accessed. It said its monitoring flagged some of the activity before Cathedral reported it, it cut off access the same day, and it disabled the affected third-party software until the vendors fixed it. Anthropic also said Cathedral did not fully disclose its testing and overstated the national-security implications.

What people are debating

Depue's core point is that big AI labs keep much of their research discussion in SaaS tools like Slack with admin-level API tokens scattered around, which sits awkwardly with their claims of being critical to national security. He also compared it to OpenAI's 2023 security incident. Supporters say supply-chain and third-party app tokens are an underrated risk for AI companies; others caution that repeating secondhand details without evidence may inflate the severity. For AI developers, the practical takeaway from the debate is that third-party apps and agent tokens connected to Slack, CRMs and similar tools need least-privilege access and regular audits.