California Attorney General Subpoenas OpenAI in Cyber Investigation

California Attorney General Rob Bonta issued an investigative subpoena to OpenAI on September 30, as part of an investigation conducted by the state's Department of Justice into incidents resulting from the company or its artificial intelligence models, including a breach that affected the Hugging Face platform.

Bonta said his office is requesting more information from the company about cyber incidents and risks, emphasizing that developers who fail to prevent cyberattacks or facilitate them "can and should be held legally accountable." This step puts the investigation on a formal track, requiring the company to provide answers to the office's questions.

An investigative subpoena is a formal request for information, not a conviction or an accusation of wrongdoing, as it is used to gather facts before deciding whether to prosecute. The move comes as part of a broader investigation announced by the state's Department of Justice into incidents resulting from AI models and their ramifications.

This action adds to a series of legal and regulatory steps targeting AI labs in the United States, which increasingly link model development to the outcomes resulting from their cyber use. The timing of the subpoena indicates that regulatory bodies have begun treating incidents related to agents and models as a matter of legal liability. It remains unclear whether the subpoena will be followed by further actions, as that depends on the information the company provides during the upcoming phase.

What do these terms mean?

Investigative subpoena: A formal request from a judicial or regulatory authority for information or documents, and is not an accusation in itself. Cyberattack: An attempt to breach or disrupt information systems, which may arise from the use of AI models. Developer accountability: Holding the company that developed the model legally responsible for the outcomes arising from its use.