Live Feeds
● LIVE Updated 1h ago · 9 sources tracked

Three Hackers Used Claude to Break Into OpenAI In Less Than 72 Hours

Three researchers breached OpenAI systems in under 72 hours using Anthropic's Claude AI model. The attack is described by some outlets as an ethical hack, while others highlight it as a significant escalation in AI-driven cybersecurity risks. The breach occurred after Anthropic released its Opus 5 model, which reportedly provided the capabilities necessary to succeed where previous versions of Claude failed. This incident demonstrates the ability of large language models to be weaponized for targeted cyberattacks against other AI developers.

🎙️

Listen to Live Briefing

Real-time synthesized voice briefing · Live Feeds Desk

⏱ ~3 min
Speed:
RSS Source map (9)
⚡ Key Developments & Real-Time Context
Text size:
  • ✓ Researchers used Anthropic's Claude models to breach OpenAI.
  • ✓ The breach was carried out by three hackers in less than 72 hours.
🛡️ Source Corroboration: 9 independent reporting domains (80% confidence) ⏱ Read time: ~2 min

What changed

Reports now specify that three hackers completed the breach in less than 72 hours using the Opus 5 model.

Live updates

  1. Three hackers use Anthropic Claude to breach OpenAI

    Three researchers breached OpenAI systems in under 72 hours using Anthropic's Claude AI model. The attack is described by some outlets as an ethical hack, while others highlight it as a significant escalation in AI-driven cybersecurity risks. The breach occurred after Anthropic released its Opus 5 model, which reportedly provided the capabilities necessary to succeed where previous versions of Claude failed. This incident demonstrates the ability of large language models to be weaponized for targeted cyberattacks against other AI developers.

    Why it matters

    The breach occurs amid heightened global concerns regarding AI safety and the security of model infrastructure. It highlights a competitive tension where one AI model is used to compromise a rival. This event raises questions about the guardrails implemented by AI labs to prevent their tools from assisting in malicious activities.

    What is confirmed

    • Researchers used Anthropic's Claude models to breach OpenAI.
    • The breach was carried out by three hackers in less than 72 hours.

    Still unconfirmed

    • The breach was an ethical hack.
    • The attack was only possible after the release of the Opus 5 model.

    What to watch next

    • Official response or post-mortem report from OpenAI regarding the breach
    • Statement from Anthropic on the specific guardrails bypassed by the hackers
    • Evidence of similar AI-assisted breaches targeting other LLM developers
    Sources used for this update (9)
    1. WSJ — Exclusive | Hackers Used Anthropic’s Claude to Break Into OpenAI
    2. Financial Times — OpenAI breached by researchers using Anthropic models
    3. Ars Technica — Researchers used Claude to hack OpenAI
    4. The Guardian — OpenAI ‘ethically hacked’ with help of Anthropic’s Claude chatbot
    5. Semafor — AI cybersecurity risks explode as Claude used to break into ChatGPT
    6. NBC News — Hackers breached OpenAI, adding to fever pitch of security and safety concerns
    7. Gizmodo — Three Hackers Used Claude to Break Into OpenAI In Less Than 72 Hours
    8. The New Stack — Claude couldn’t hack OpenAI. Then Anthropic shipped Opus 5.
    9. www.tweaktown.com — Google Gemini broke out of its test environment and hacked three real companies
    confidence 80%
📊

Community Sentiment: How do you assess this situation?

Voice your perspective · Real-time aggregated sentiment from the Live Feeds community