Matthew Green: 5 Months On, No Evidence of a Security Team Covering AI Research Org

matthew_d_green · x · 2026-09-30

Security researcher Matthew Green says that as of nearly October 2026, he sees no external evidence that any frontier AI company has built a security group focused on protecting the research org. The CISO and security team were focused on product security, while incidents happened on the research side — in Spring 2026 nobody assumed infra security needed to cover research.

In a follow-up he adds that, watching the CEO personally handle incident response, any security professional would conclude 'this operation is a clown show.' It's entirely fixable, he writes, but five months on it remains unaddressed.

Related event: OpenAI Faces Questions Over Agent Safety Communication and Research-Side Security Gaps(5 posts)→

Original post →

More from Companies & People

Companies & People channel →