Security
Part of The AI Control Crisis · 15 stories · since Sep 3 · newest 11m ago
Claude Opus 5 used to breach OpenAI's internal systems in under 72 hours
OpenAI confirms breach and patches both vulnerabilities
2
Sep 18
· 5d ago · 2 articles · 1 post · 2 sources · development 2 of 3
OpenAI acknowledged two security issues—one in Discourse and another in OpenAI's own systems—and stated both are now resolved. The company narrowed permissions on community sign-in tokens and revoked affected sessions.
nypost.com
“We thank the researchers for contacting us and sharing their findings. We narrowed the permissions on Community sign-in tokens and revoked affected tokens and sessions.”
OpenAI Mohan Pedhapati Chief Technology Officer, Hacktron AI Hacktron AI Security research firm OpenAI AI company attacked Anthropic AI company whose Claude was used in breach
The whole story the bright band is this development · numbered dots are the others · click one to jump
Peak 12 pieces in two hours at Sep 17, 11 PM; 69 pieces over 7 days (36 articles · 1 video · 32 posts)
Sep 16, 1 PM — 1 piece · 1 article — Google News 1 Sep 16, 3 PM — quiet Sep 16, 5 PM — quiet Sep 16, 7 PM — quiet Sep 16, 9 PM — quiet Sep 16, 11 PM — quiet Sep 17, 1 AM — quiet Sep 17, 3 AM — quiet Sep 17, 5 AM — quiet Sep 17, 7 AM — quiet Sep 17, 9 AM — quiet Sep 17, 11 AM — quiet Sep 17, 1 PM — quiet Sep 17, 3 PM — quiet Sep 17, 5 PM — quiet Sep 17, 7 PM — quiet Sep 17, 9 PM — 1 piece · 1 article — Google News 1 Sep 17, 11 PM — 12 pieces · 10 articles · 2 posts — Newswires 10, Bluesky 1, Mastodon 1 Sep 18, 1 AM — quiet Sep 18, 3 AM — quiet Sep 18, 5 AM — 1 piece · 1 article — Newswires 1 Sep 18, 7 AM — 1 piece · 1 post — Mastodon 1 Sep 18, 9 AM — 6 pieces · 6 articles — Newswires 3, Google News 3 Sep 18, 11 AM — 4 pieces · 2 articles · 2 posts — Newswires 2, Mastodon 1, Reddit 1 Sep 18, 1 PM — 5 pieces · 2 articles · 3 posts — Mastodon 2, Newswires 1, Google News 1, +1 more Sep 18, 3 PM — 2 pieces · 2 posts — Mastodon 2 Sep 18, 5 PM — 5 pieces · 5 articles — Google News 3, Newswires 2 Sep 18, 7 PM — 3 pieces · 2 articles · 1 video — Newswires 2, YouTube 1 Sep 18, 9 PM — quiet Sep 18, 11 PM — quiet Sep 19, 1 AM — 4 pieces · 1 article · 3 posts — Hacker News 1, Newswires 1, Bluesky 1, +1 more Sep 19, 3 AM — quiet Sep 19, 5 AM — quiet Sep 19, 7 AM — 3 pieces · 2 articles · 1 post — Hacker News 1, Google News 1, Newswires 1 Sep 19, 9 AM — quiet Sep 19, 11 AM — 1 piece · 1 article — Newswires 1 Sep 19, 1 PM — 1 piece · 1 post — Bluesky 1 Sep 19, 3 PM — 5 pieces · 1 article · 4 posts — Mastodon 3, Hacker News 1, Newswires 1 Sep 19, 5 PM — 1 piece · 1 post — Bluesky 1 Sep 19, 7 PM — quiet Sep 19, 9 PM — 1 piece · 1 article — Google News 1 Sep 19, 11 PM — quiet Sep 20, 1 AM — 1 piece · 1 post — Bluesky 1 Sep 20, 3 AM — quiet Sep 20, 5 AM — quiet Sep 20, 7 AM — 1 piece · 1 post — Mastodon 1 Sep 20, 9 AM — quiet Sep 20, 11 AM — quiet Sep 20, 1 PM — 2 pieces · 2 posts — Bluesky 2 Sep 20, 3 PM — quiet Sep 20, 5 PM — quiet Sep 20, 7 PM — quiet Sep 20, 9 PM — quiet Sep 20, 11 PM — quiet Sep 21, 1 AM — quiet Sep 21, 3 AM — quiet Sep 21, 5 AM — quiet Sep 21, 7 AM — quiet Sep 21, 9 AM — 1 piece · 1 post — Bluesky 1 Sep 21, 11 AM — 3 pieces · 3 posts — Bluesky 2, Mastodon 1 Sep 21, 1 PM — 2 pieces · 2 posts — Hacker News 1, Mastodon 1 Sep 21, 3 PM — quiet Sep 21, 5 PM — quiet Sep 21, 7 PM — quiet Sep 21, 9 PM — quiet Sep 21, 11 PM — quiet Yesterday, 1 AM — quiet Yesterday, 3 AM — 1 piece · 1 post — Bluesky 1 Yesterday, 5 AM — quiet Yesterday, 7 AM — quiet Yesterday, 9 AM — quiet Yesterday, 11 AM — quiet Yesterday, 1 PM — quiet Yesterday, 3 PM — 1 piece · 1 post — Bluesky 1 Yesterday, 5 PM — quiet Yesterday, 7 PM — quiet Yesterday, 9 PM — quiet Yesterday, 11 PM — quiet Today, 1 AM — quiet Today, 3 AM — quiet Today, 5 AM — quiet Today, 7 AM — quiet Today, 9 AM — quiet Today, 11 AM — quiet Today, 1 PM — quiet Today, 3 PM — quiet Today, 5 PM — quiet Today, 7 PM — quiet Today, 9 PM — quiet 1 2 3
Sep 17 Sep 18 Sep 19 Sep 20 Sep 21 yesterday now · 11:58 PM ET
Reported in the same hours
no headline names this development itself — these 1 claim were published in its stretch
first by Fortune , 5d ago
What people said 1 voice · verbatim
H
A three-person cybersecurity team used Anthropic’s Claude to help hack OpenAI during authorized research, taking over employee accounts and reaching an internal code repository in under 72 hours. The reward? A $6,500 bug bounty. Listen/Read: https:// hackread.com/cybersecurity-sta rtup-claude-ai-hack-openai-bug-bounty/ # Cybersecurity # OpenAI #…
All 3 developments of Claude Opus 5 used to breach OpenAI's internal systems in… →
H