Researchers used Claude to hack OpenAI
Summary
Cybersecurity researchers used a security tool from AI company Anthropic to find a weakness in OpenAI’s system. They accessed an OpenAI employee’s ChatGPT account and internal code but reported the problem and were paid to help fix it.Key Facts
- A small cybersecurity group called Hacktron AI accessed an OpenAI employee’s ChatGPT account using a tool from Anthropic.
- They found a security flaw involving OpenAI’s community forum platform, which was hosted by a third party named Discourse.
- The accessed ChatGPT account had permission to view OpenAI’s internal software code on GitHub.
- OpenAI paid the researchers $6,500 under a bug bounty program, where companies pay hackers to find security weaknesses responsibly.
- OpenAI fixed the security issue after the researchers reported it.
- This event raises concerns about the safety of AI systems as these models become more powerful.
- Anthropic released data showing that 26% of their research is now led by their AI model, Claude, up from 1% earlier in the year.
- Anthropic explained that AI is increasingly used to improve itself but still works mostly under human supervision.
Read the Full Article
This is a fact-based summary from The Actual News. Click below to read the complete story directly from the original source.
