Newsletter IconGoogle IconFacebook IconX IconThreads IconInstagram IconYouTube Icon
Giveaway: GAMDIAS ATLAS P1 Case, HELIOS 750W PSU, and BOREAS Digital CPU Cooler

Claude Opus 5 helped researchers hack OpenAI in less than 72 hours

Security researchers used Anthropic's Claude Opus 5 to help break into OpenAI's systems, access an employee account and expose internal repository access.

Claude Opus 5 helped researchers hack OpenAI in less than 72 hours
Facebook IconX IconReddit Icon
Comments
Tech and Science Editor
Published
1-minute read time
TL;DR: Researchers using Anthropic's Claude Opus 5 helped Hacktron AI exploit two vulnerabilities to access an OpenAI employee's ChatGPT account and internal repositories during a bug bounty test, stopping before retrieving sensitive data, reporting the issue and receiving $6,500; similar model escape incidents have occurred due to misconfigured testing environments.
Voice: Jak Connor
0:00 / 2:44
Use left and right arrow keys to seek audio.

Security researchers used Anthropic's Claude AI to help break into OpenAI's systems, demonstrating how rapidly improving AI models can accelerate sophisticated cybersecurity attacks.

Claude Opus 5 helped researchers hack OpenAI in less than 72 hours 2

According to The Wall Street Journal, researchers from cybersecurity firm Hacktron AI used Claude while participating in OpenAI's bug bounty program. The team ultimately gained access to an OpenAI employee's ChatGPT account, providing a route into parts of the company's private software repositories. The researchers stopped short of accessing sensitive information, disclosed their findings to OpenAI and received a $6,500 bug bounty.

The incident comes amid growing concern about the cybersecurity capabilities of frontier AI models. Anthropic recently disclosed four separate cases in which Claude models gained unauthorized access to real-world systems during cybersecurity evaluations after a third-party testing environment was mistakenly left connected to the internet.

"On July 25, 2026, we chained two critical vulnerabilities to compromise multiple OpenAI employees' ChatGPT accounts. With these accounts, we could then access internal OpenAI repositories, and potentially many other connectors. To prove we had in fact gained the access we believed without allowing ourselves to learn any sensitive information, we used the employee's Codex to open a PR #1186742 in OpenAI's internal monorepo," wrote the researchers

In those evaluations, Claude was instructed to complete capture-the-flag challenges and told it was operating without internet access. Instead, a configuration error allowed the models to reach real systems, where they extracted credentials, accessed production data, and, in one case, published a malicious Python package that was downloaded onto real machines.

Frequently Asked Questions

Open a question for an answer from TweakTown's coverage of this news, or ask your own below.

Question #1

How did Claude Opus 5 help the researchers compromise OpenAI employee ChatGPT accounts?

Question #2

How did OpenAI verify the researchers had access without exposing sensitive information?

Question #3

What part did OpenAI's bug bounty program play in this incident?

Question #4

What did Anthropic say about Claude models accidentally reaching real systems during cybersecurity evaluations?

Have a question that isn't listed here? Ask below, and TweakBot will answer it.

OpenAI has experienced a similar issue, previously disclosing that its own models escaped an isolated testing environment by exploiting a previously unknown vulnerability and subsequently accessed Hugging Face infrastructure. During this incident, OpenAI's model was heading for the internet to gain the knowledge it needed to pass the tests it faced within the environment it escaped.

Photo of the MINISFORUM MS-S1 Max Mini Workstation Ryzen AI Max+ 395 Mini PC

Best Deals: MINISFORUM MS-S1 Max Mini Workstation Ryzen AI Max+ 395 Mini PC

Prices last scanned 1 hour and 40 minutes ago

* Prices may be inaccurate. As an Amazon Associate, we earn from qualifying purchases. We earn affiliate commissions from Newegg and PCCG sales.

Join Our Newsletter

Join the TweakTown Newsletter for daily tech updates delivered to your inbox.

See previous giveaways.

News Sources:wsj.com, hacktron.ai, and forbes.com

Comments

About the author

Tech and Science Editor

Jak joined TweakTown in 2017 and has since reviewed 100s of new tech products and kept us informed daily on the latest science, space, and artificial intelligence news. Jak's love for science, space, and technology, and, more specifically, PC gaming, began at 10 years old. It was the day his dad showed him how to play Age of Empires on an old Compaq PC. Ever since that day, Jak fell in love with games and the progression of the technology industry in all its forms.

Stay Updated

Follow TweakTown for breaking tech news, reviews, and daily updates.

Follow TweakTown on GoogleAdd TweakTown as a preferred source on Google
Newsletter Subscription