Independent security researchers used Anthropic’s Claude to break into an OpenAI employee’s account and reach the company’s internal code. It follows weeks after an experiment in which artificial intelligence (AI) agents broke out of containment at ChatGPT maker OpenAI to hack the company Hugging Face.
Cybersecurity Firm Hacks OpenAI Using Anthropic’s Claude AI
Researchers at Hacktron AI used Anthropic’s Claude models to hack and gain access to an OpenAI employee’s ChatGPT and Codex accounts. The security researchers gained a path into internal repositories.
According to founder s1r1us, the team hacked into OpenAI using two bugs in ChatGPT or Codex accounts. Successfully managed to reach connected services including Outlook, Slack, GitHub, and others.
“We proved it with a PR in OpenAI’s internal codebase. It took us <72 hrs,” said s1r1us. The team then stopped testing and reported the issues to Discourse and OpenAI.
Meanwhile, OpenAI has also confirmed a fix within about 14 hours and paid a $6,500 bounty. This was part of a bug bounty program where researchers were given access to Anthropic’s Claude models.
Claude played a central role in exploit development using Claude Opus 4.8 and Opus 5 models. The team ran models in an autonomous loop against a test instance before applying the generated script to OpenAI’s forum.
The researchers claimed AI is lowering the cost of turning known or obscure memory-corruption bugs into reliable exploits. They noted that similar libheif issues affect a wide range of software that processes HEIC/HEIF images.
They argued that “even leading AI labs can be vulnerable. Defenders need to fix the architecture, patch faster, and limit the blast radius of connected things.”
Our main takeaway from hacking OpenAI: AI is reducing the amount of scarce expertise needed to develop exploits. Work that once took months can now take days.
Even leading AI labs can be vulnerable.
Defenders need to fix the architecture, patch faster, and limit the blast…
— s1r1us (@S1r1u5_) September 18, 2026
Warning After ChatGPT Maker Hugging Face Hack Experiment
OpenAI revealed earlier that almost 700 out of 1200 artificial intelligence models coordinated an attack and two models broke out of a sandboxed testing environment to hack into the machine learning platform Hugging Face.
AI executives urged a slowdown in development and push for more safety rules, shaking investors’ confidence in the industry. Anthropic CEO Dario Amodei’s essay “We Must Pace the Frontier” raised warnings about AI models helping build the next generation of models.
OpenAI’s Sam Altman and xAI’s Elon Musk agreed on rising AI risks to humanity and the need for AI safety. Google DeepMind AGI safety researcher Bilal Chughtai resigned, warning AI “has the potential to kill us all.”
Meanwhile, Ethereum co-founder Vitalik Buterin rejected AI hacking can break cybersecurity and crypto claims. However, he believes it is necessary to verify the entire program amid modern AI risks. Vitalik Buterin emphasized that developers and security teams need to move fast to have AI on the cybersecurity side.
Also Check: Best Crypto AI Trading Bots in 2026: Which Bots Actually Use AI?
