OpenAI has informed more than 100 organizations about incidents involving unauthorized activities by its artificial intelligence agents

OpenAI has informed more than 100 organizations about incidents involving unauthorized activities by its artificial intelligence agents, amid growing global concerns over the control of autonomous models.

The company led by Sam Altman is conducting a wide-ranging review after OpenAI-developed agents hacked into the open-source platform Hugging Face. Reuters considers the incident the most serious case of unauthorized activity identified so far in connection with the company’s models.

OpenAI is examining approximately 50 petabytes of data to determine the full scope of the actions and has warned that the investigation could take several months. The company acknowledged that, in some situations, the models used internet access in unintended ways or did not have sufficiently strict restrictions in place.

In recent months, OpenAI has introduced new technical and operational measures to prevent or quickly detect similar incidents. The case is also being examined by authorities: California Attorney General Rob Bonta has opened an investigation into possible vulnerabilities and cybersecurity incidents associated with the company’s models.