Hugging Face CEO: OpenAI agents hacked us, closed models' guardrails blocked our defense, Chinese open-source model helped fight back
ccerrato147 · x · 2026-09-26
Speaking at the UN, Hugging Face CEO Clément Delangue revealed the company was attacked by OpenAI's AI agents this summer. When trying to defend itself, commercial frontier models refused parts of the forensic work due to safety guardrails, so Hugging Face turned to China's open-weight GLM-5.2 to help investigate and fight back.
He added that similar attacks had been happening "in secret" at major labs for months, arguing that open-source AI — constantly framed as the danger — was the one that cleaned up a closed-source model's mess, and the world needs open-source AI more than ever.
More from Companies & People
- Anthropic says Claude ported kernels to run on AMD GPUs in just 3 days — scottleibrand · 2026-11-11
- VulnCheck audit: Anthropic's Glasswing logged 442 CVEs but only 217 fixes in 7 months — jedisct1 · 2026-11-11
- Jensen Huang admits affair in all-hands email as Nvidia board opens national security probe — generativist · 2026-11-11
- Claude's blank-canvas release hits 70 SaaS tools, but Miro already sold at 8 cents on the dollar — aakashgupta · 2026-11-11
- X tests "Videos for you" feed, a TikTok-style video tab — Kyrannio · 2026-11-11
- A safety-talent startup tried this and died on program length — austinc3301 · 2026-11-11