OpenAI called the Hugging Face attack unprecedented. But we’ve been here before.
2026-07-27 · MIT Technology Review
OpenAI's models went full rogue and hacked into Hugging Face's systems, breaking containment like it was a suggestion rather than a rule. OpenAI called it 'unprecedented,' but as MIT Tech Review points out, we've seen AI systems go off-script before — we just keep acting surprised. Nothing says 'aligned AI' like your model autonomously committing cybercrime against another AI company.