Cybersecurity researchers from the US-based startup Hacktron AI have successfully compromised several OpenAI employee accounts, an operation they claim provided them with potential access to a vast array of software caches. The team utilized Anthropic’s Claude chatbot to generate code, which facilitated unauthorized access to ChatGPT accounts through an OpenAI staff discussion forum hosted on the Discourse platform. Following this initial breach, the researchers executed a harmless “pull request” to modify code within OpenAI’s service on the GitHub software repository.
Despite leveraging Claude during the early stages of the attack, the researchers noted that they primarily utilized OpenAI’s own advanced GPT-5.6 Sol model to conduct the hack. Hacktron AI reported their findings to OpenAI under the company’s official bug bounty programme, which compensates ethical hackers for identifying and testing system vulnerabilities. The researchers emphasized that while they possessed the capability to download code from the GitHub repository, they refrained from doing so. OpenAI has confirmed that it has since addressed the vulnerabilities exploited during the operation.
According to Hacktron, the integration of AI tools has significantly reduced the time and resources required to plan and execute complex cyberattacks. The firm stated that tasks which previously demanded months of effort and a well-resourced team can now be completed in just a few days. For their contributions to improving system security, Hacktron received a $6,500 payment from OpenAI.
This incident occurs amid broader industry concerns regarding the rapid progression of artificial intelligence. OpenAI recently disclosed six more examples of “unexpected or concerning” behavior from its technology, warning that current development speeds may not be sustainable. Meanwhile, Anthropic has renewed calls for a slowdown in AI development, a stance supported by Google DeepMind, OpenAI, and Elon Musk, who argue that unrestrained progress poses an existential threat. In contrast, Donald Trump has rejected these calls for a pause, citing the necessity of maintaining a competitive advantage over China’s AI industry.
