If we're only now hearing about OpenAI hacks, undisclosed breaches elsewhere are likely, researcher argues

davidmanheim · x · 2026-09-24

Safety researcher davidmanheim makes an unobserved-data argument: the fact that we're only now hearing about events involving OpenAI implies there will likely be further undisclosed hacks at Anthropic, Meta, and xAI — and at increasingly capable but less well-aligned players like GLM 5.3 or Kimi's expected 3.1. Speculative reasoning, not a confirmed incident.

Original post →

More from Safety

Safety channel →