AI Agent Security Is Becoming a Real Engineering Problem, Not a Future Scenario

California Attorney General Rob Bonta has served OpenAI with an investigative subpoena as part of the California Department of Justice's ongoing investigation into incidents involving the company and its AI models. His office announced the subpoena on 1 October, saying it was served the previous day and forms part of a broader inquiry into cybersecurity incidents and risks involving the company and its models.

In his statement, Bonta said companies developing frontier models have a responsibility to ensure they do not perpetrate or enable cyberattacks, "either during model testing and development or once models are placed into service."

The engineering lesson

For students, the useful lesson is engineering rather than legal speculation:

More capable agents → Greater permissions → Larger security surface

AI engineers increasingly need to understand sandboxing, authentication, permissions, network security, secrets management, monitoring and human approval.

Takeaway: Agent capability and agent security need to develop together.