Hacktron AI uses Anthropic's Claude and OpenAI's GPT model to breach employee ChatGPT accounts
A US startup's security team accessed OpenAI staff ChatGPT accounts and a GitHub repository using AI tools, reporting the vulnerabilities through OpenAI's bug bounty program.
A team of researchers at Hacktron AI successfully compromised several OpenAI employees' ChatGPT accounts by exploiting a discussion forum hosted on the Discourse platform. The breach allowed the team to reach the users' software cache and submit a harmless pull request to OpenAI's GitHub repository.
The operation began with Anthropic's Claude chatbot, which generated code to facilitate the intrusion, but the researchers said most of the work was carried out using OpenAI's own GPT-5.6 Sol model. Hacktron reported the findings to OpenAI under the company's bug bounty program and received a $6,500 reward.
OpenAI acknowledged the report, thanking the researchers and confirming that the identified vulnerabilities had been patched.
Hacktron noted that AI tools reduced the time required for the attack from months to days, a trend highlighted by cybersecurity experts as a growing concern. The incident follows previous safety issues at OpenAI, including a recent test in which autonomous AI agents accessed the Hugging Face platform.
Anthropic, along with other AI leaders, has called for a slowdown in AI development, citing potential risks, while some political figures have dismissed the calls.