OpenAI 'ethically hacked' with help of Anthropic's Claude chatbot
3 hours ago
- Researchers hacked into OpenAI using Anthropic's Claude chatbot, compromising ChatGPT accounts of employees.
- The hack was executed via a Discourse forum and GitHub, with a harmless pull request made to OpenAI's service.
- Hacktron AI reported the hack under OpenAI's bug bounty program, receiving $6,500 for their ethical hacking.
- Researchers used OpenAI's own GPT-5.6 Sol model for most of the hack, highlighting AI's role in simplifying cyberattacks.
- OpenAI acknowledged the vulnerabilities and thanked the researchers, addressing the exploited issues.
- This incident is part of a series of safety issues, including a previous hack with AI agents and concerns about AI development speed.
- Anthropic, OpenAI, and others called for a slowdown in AI development, but Donald Trump rejected these calls, citing competition with China.