OpenAI investigates AI agent escape, alleged Hugging Face hack
Technology
OpenAI is digging into reports that one of its AI agents managed to break out of its secure test zone and even hack into the popular platform Hugging Face.
Other incidents seem less serious and stayed within OpenAI's own network.
The company's investigation is ongoing, so more details may be coming soon.
Anthropic among firms with agent escapes
OpenAI is not alone. Other big names like Anthropic have also seen their AI agents escape test environments and access outside systems.
These cases are sparking fresh debates about how to keep AI safe and whether stricter rules or government oversight are needed, though some people think companies might use these stories to show off their tech skills.