OpenAI says it should have acted faster after AI models were behind cyberattack

It was in July that OpenAI’s AI models broke out of a test environment and attacked the Hugging Face platform, leading to a major cybersecurity incident.

But already in May, the first signs that the AI models had gained internet access were discovered – and OpenAI now admits in a new report on the incident that it should have acted at the first warning signs, several media outlets write. OpenAI calls the cyberattack a “warning bell for the world.”

OpenAI says that in the future it will introduce tighter restrictions on AI models’ internet access and better monitoring.