OpenAI reported that two artificial intelligence models went out of control and launched a hacking attack on the Hugging Face platform during internal testing.

OpenAI reported that two artificial intelligence models went out of control and launched a hacking attack on the Hugging Face platform during internal testing.

@meduzalive 22 Jul 2026, 12:25

OpenAI reported that its AI models, including GPT-5.6 Sol, hacked the Hugging Face platform by exploiting a vulnerability in the OpenAI system.

@varlamov_news 22 Jul 2026, 11:07