conv.

All stories
SecurityQuiet 5d · day 6

Researchers use Claude to breach ChatGPT in under 72 hours

Security researchers demonstrated AI-powered hacking by using Anthropic's Claude to penetrate OpenAI systems, highlighting urgent vulnerabilities in leading language models.

newest coverage 4d agosummary written 5d ago6 pieces NewswiresMastodonGoogle News

What to know

  • Researchers breached ChatGPT in under 72 hours using Claude, accessing an employee account, source code repository details, and forums—then responsibly reported the vulnerability.
  • OpenAI patched the flaw and paid a $6,500 bounty; the breach was part of a legitimate bug-hunting program.
  • Security experts argue the real near-term risk is insufficient technical guardrails, not superintelligent AI—and warn AI-powered attacks threaten critical infrastructure.
  • The breach joins recent incidents like the July Hugging Face hack, prompting calls from AI leaders including Anthropic's CEO for the industry to slow development pace.

“We immediately reported the initial vulnerability to OpenAI and Discourse and worked with them to coordinate the patch. We appreciate their attention to detail and fast resolution of this issue.”

Hacktron AI, Security research platform · CBS News ↗

Hacktron AI Security research platformOpenAI ChatGPT developerAnthropic Claude AI developerDario AmodeiDario Amodei Anthropic CEO

Researchers use Claude to breach ChatGPT in under 72 hours
Semafor

How it unfolded 2 developments, newest first · click a bar or a number to jump articlesposts

Peak 2 pieces in two hours at Sep 17, 11 PM; 6 pieces over 6 days (5 articles · 1 post) Sep 17, 11 PM — 2 pieces · 2 articles — Newswires 2Sep 18, 1 AM — quietSep 18, 3 AM — quietSep 18, 5 AM — 1 piece · 1 article — Newswires 1Sep 18, 7 AM — quietSep 18, 9 AM — 1 piece · 1 article — Newswires 1Sep 18, 11 AM — quietSep 18, 1 PM — 1 piece · 1 post — Mastodon 1Sep 18, 3 PM — quietSep 18, 5 PM — quietSep 18, 7 PM — quietSep 18, 9 PM — quietSep 18, 11 PM — quietSep 19, 1 AM — quietSep 19, 3 AM — quietSep 19, 5 AM — 1 piece · 1 article — Google News 1Sep 19, 7 AM — quietSep 19, 9 AM — quietSep 19, 11 AM — quietSep 19, 1 PM — quietSep 19, 3 PM — quietSep 19, 5 PM — quietSep 19, 7 PM — quietSep 19, 9 PM — quietSep 19, 11 PM — quietSep 20, 1 AM — quietSep 20, 3 AM — quietSep 20, 5 AM — quietSep 20, 7 AM — quietSep 20, 9 AM — quietSep 20, 11 AM — quietSep 20, 1 PM — quietSep 20, 3 PM — quietSep 20, 5 PM — quietSep 20, 7 PM — quietSep 20, 9 PM — quietSep 20, 11 PM — quietSep 21, 1 AM — quietSep 21, 3 AM — quietSep 21, 5 AM — quietSep 21, 7 AM — quietSep 21, 9 AM — quietSep 21, 11 AM — quietSep 21, 1 PM — quietSep 21, 3 PM — quietSep 21, 5 PM — quietSep 21, 7 PM — quietSep 21, 9 PM — quietSep 21, 11 PM — quietSep 22, 1 AM — quietSep 22, 3 AM — quietSep 22, 5 AM — quietSep 22, 7 AM — quietSep 22, 9 AM — quietSep 22, 11 AM — quietSep 22, 1 PM — quietSep 22, 3 PM — quietSep 22, 5 PM — quietSep 22, 7 PM — quietSep 22, 9 PM — quietSep 22, 11 PM — quietYesterday, 1 AM — quietYesterday, 3 AM — quietYesterday, 5 AM — quietYesterday, 7 AM — quietYesterday, 9 AM — quietYesterday, 11 AM — quietYesterday, 1 PM — quietYesterday, 3 PM — quietYesterday, 5 PM — quietYesterday, 7 PM — quietYesterday, 9 PM — quietYesterday, 11 PM — quietToday, 1 AM — quietToday, 3 AM — quiet 1–2
Sep 18Sep 19Sep 20Sep 21Sep 22yesterdaynow · 4:44 AM ET
  1. 2

    OpenAI patches vulnerability and pays bounty

    OpenAI responded promptly to the reported breach, narrowed permissions on Community sign-in tokens, revoked affected tokens and sessions, and paid the researchers a $6,500 bounty.

    “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

    • AI cybersecurity risks explode as Anthropic's Claude used to break into OpenAI's ChatGPT systems

      semafor@threads.netMastodon5d agoview on Mastodon ↗
  2. background

    Hacktron AI researchers breach ChatGPT using Claude — Security researchers from Hacktron AI used Anthropic's Claude to gain access to an OpenAI employee's ChatGPT account, retrieve source code repository details, and access an OpenAI discussion forum. The entire process took less than 72 hours.

  3. 1

    Experts attribute breaches to guardrail gaps, not superintelligence

    Analysis indicates recent breaches result from insufficient technical guardrails leaving networks vulnerable, rather than from developing superintelligence as some AI leaders warn. Business executives and ex-government officials voice concern that AI-powered cyberattacks could threaten critical infrastructure, while the Pentagon's outdated computer networks have exacerbated cybersecurity risks.

    “The entire timeline from initial discovery to access to OpenAI repo access took place in less than 72 hours.”
    — Hacktron AI, Security research platform · source
    1. 1 outlet AI cybersecurity risks explode as Claude used to break into ChatGPT

      first by Semafor, 5d ago