# Security researchers used Claude to help hack into OpenAI

https://mins.news/en/item/64492

- The Verge, 2026-09-18 15:30 UTC
- Tech
- Original: https://www.theverge.com/ai-artificial-intelligence/997444/openai-hack-claude-heif-heist

> Three independent Hacktron researchers say they compromised OpenAI employee accounts in under 72 hours with Anthropic’s Claude Opus 4.8 and 5. They reached OpenAI’s GitHub “Monorepo” through a HEIF-processing flaw in Discourse, but did not access its internal code. A Codex pull request demonstrated the breach. The vulnerabilities were fixed, and OpenAI paid Hacktron $6,500.

## Stories

- [AI Cybersecurity and Loss of Control Risks](https://mins.news/en/topic/ai-cybersecurity-and-control) — Markdown: https://mins.news/en/topic/ai-cybersecurity-and-control.md

## Related coverage

- [Researchers used Claude to breach OpenAI’s internal systems](https://mins.news/en/item/65146) — CGTN, 2026-09-19
- [Anthropic’s Claude helped researchers breach OpenAI in under 72 hours](https://mins.news/en/item/65060) — CryptoSlate, 2026-09-19
- [Report: Researchers breached OpenAI with Claude](https://mins.news/en/video/64318) — ABC News, 2026-09-18