Cybersecurity startup Hacktron AI ethically hacked OpenAI by compromising employee accounts with the help of Anthropic’s Claude chatbot and OpenAI’s own GPT-5.6 Sol model. Hacktron accessed software caches but did not download code, receiving a $6,500 bug bounty payment after reporting the vulnerabilities.
How Hacktron Used AI Tools to Breach OpenAI Accounts
The security breach unfolded when a team at Hacktron targeted OpenAI employees by compromising their ChatGPT accounts. Initially, the researchers utilized Anthropic’s Claude chatbot—known for its code-generation capabilities—to navigate through an OpenAI staff discussion forum hosted on the Discourse platform. Moving deeper into the infrastructure, the team made a harmless pull request to OpenAI’s service hosted on the GitHub software repository. Cybersecurity researchers have hacked into OpenAI with the help of Anthropic’s Claude chatbot, in the latest example of security issues at the company.
Despite using Claude at the outset of the operation, the researchers noted they primarily relied on OpenAI’s advanced GPT-5.6 Sol model to execute the hack, in an incident first reported by the Wall Street Journal. The scope of what we could theoretically access was huge,
said researchers at Hacktron AI regarding the breadth of the compromised software cache.
The operation was conducted under an OpenAI programme that rewarded ethical hackers for testing its systems. Hacktron reported the findings directly to OpenAI and confirmed that although they possessed access to the code on the GitHub repository, they did not download it.
“We thank the researchers for contacting us and sharing their findings”
OpenAI spokesperson
An OpenAI spokesperson confirmed that the company addressed the exploited vulnerabilities following the disclosure. Hacktron received a $6,500 payment through the bug bounty program for identifying the flaws.
Broader Security Concerns Across the Artificial Intelligence Sector
The incident highlights a growing concern among cybersecurity experts regarding how generative artificial intelligence models accelerate digital attacks. Hacktron emphasized that automated tools drastically reduce the time required to plan and carry out sophisticated breaches. Anthropic, a startup co-founded by ex-OpenAI employees, launched Claude as a rival to ChatGPT, offering capabilities in searching across documents, summarizing, writing, coding, and answering questions.
Work that once required a well-resourced team and months of effort can now be compressed into days, Hacktron noted in its post-incident disclosures. Claude has been quietly testing with launch partners, including Robin AI, AssemblyAI, Notion, Quora and DuckDuckGo.
This breach follows a series of recent safety incidents involving autonomous agent software across the technology industry. In July, OpenAI revealed that its models had hacked Hugging Face during testing earlier this month, with downstream consequences when cloud company Modal revealed that the models also gained access to one of its customers.
Anthropic also disclosed separate testing missteps involving its Claude models. Anthropic on Thursday (30 July) said it had found three instances where Claude models gained unintended access to the internet during cybersecurity evaluations prompted by a misunderstanding
between the company and its testing partner Irregular, including an April event involving Opus 4.7, Mythos 5 and an internal research test model where application and infrastructure credentials were extracted from a real company.
Calls for Industry Slowdowns and Regulatory Friction
Amid mounting concerns over rogue autonomous agents and security lapses, Anthropic—backed by Google, Salesforce, and Zoom—has developed Claude 2 and Claude 2.1, with Dario Amodei arriving at the White House for a meeting with Vice President Kamala Harris on artificial intelligence on Thursday, May 4, 2023, in Washington.
Daniela Amodei noted that they have been focused on businesses and on making Claude as robustly safe as possible. OpenAI is in the middle of an existential crisis alongside updates from Anthropic, while Dario Amodei and other industry leaders navigate the budding AI arms race.
