Security Researchers Breach OpenAI Internal Systems via Claude
A three-person team claims it used Anthropic's Claude model to access an OpenAI employee account and a GitHub repository said to contain 'OpenAI's algorithmic secrets' in under 72 hours.
Security Researchers Breach OpenAI Internal Systems via Claude
According to The Wall Street Journal, a team of three independent security researchers operating under the name Hacktron claims to have breached an OpenAI employee account in under 72 hours using Anthropic's Claude model. The team reportedly gained access to OpenAI's GitHub code repository known as "Monorepo," which people familiar with the matter say contains "OpenAI's algorithmic secrets."
The researchers did not directly read the code inside Monorepo. Instead, they submitted a pull request through an employee's Codex account to prove they had obtained access. The entry point for the intrusion was Discourse, a third-party forum service.
The incident brings the real-world risk of AI models being used to assist cyberattacks to the forefront, and puts the boundaries of model capabilities and access controls in the spotlight. As of publication, neither OpenAI nor Anthropic has publicly responded.
Source: https://www.theverge.com/ai-artificial-intelligence/997444/openai-hack-claude-heif-heist
Provenance & status
- Byline
- OceanAlt Editorial
- First published
- 2026-09-19
- Last updated
- 2026-09-30
- Content type
- Newsflash
- Source material
- View original ↗
Related reading
Paste a payee address before you pay and see whether it's on a sanctions list, through a mixer, or tagged for fraud.
This judgement can sit inside your own product
One line of code; it touches neither your CSS nor your JS. The same pre-settlement judgement can appear in your articles, on your wallet's confirmation screen, or as an endpoint your agent calls before it pays.