Red open padlock icons over blue digital circuits and binary code illustrate cybersecurity vulnerabilities.

California Sues OpenAI Over AI Security Risks

The lawsuit alleges inadequate testing safeguards allowed autonomous AI agents to breach outside networks and endanger the public.

California Attorney General Rob Bonta sued OpenAI in San Francisco Superior Court on Monday, alleging the company misled the public about its security controls while testing advanced artificial intelligence models without adequate safeguards.

According to the California Department of Justice’s announcement, the lawsuit alleges OpenAI created a public nuisance and violated California’s Unfair Competition Law by failing to properly secure and monitor its model testing.

The legal action follows a September investigation into a July 2026 cyberattack on open-source AI hosting platform Hugging Face. State prosecutors said hundreds of autonomous AI agents coordinated the attack over more than two days after escaping OpenAI’s research environment, one of at least 80 times OpenAI admitted to losing control of autonomous agents during testing.

Bonta’s office alleges OpenAI pressed ahead with experiments despite evidence that its safeguards could not contain the models and missed or disregarded signs that agents were misusing outside organizations’ resources. Subsequent incidents allegedly involved unauthorized access to software code repositories, informational websites, government databases storing sensitive health information and OpenAI’s own internal infrastructure.

The state is asking the court to order changes to OpenAI’s business practices and require the company to give up profits tied to the alleged misconduct, The Associated Press reported.

OpenAI spokesperson Drew Pusateri told the AP that the lawsuit appears to single out OpenAI over an industrywide challenge. He said the company has strengthened safeguards, published technical and independent reports and created a framework to detect when AI agents act contrary to instructions or intended goals, adding that OpenAI continues to work with Bonta’s office on industry safety standards.

Security Today has contacted OpenAI for additional comment.

About the Author

Chelsey Arries is assistant editor of Security Today, Campus Security Today, Occupational Health & Safety and Environmental Protection Online.

Featured

New Products

  • Schlage Sense Pro

    Schlage Sense Pro™ Smart Deadbolt

    Deliver true hands-free entry and reliable smart home integration with a premium deadbolt featuring Schlage Converge™ technology and native Matter over Thread support.

  • Adams Rite RITE Touch® RT2050 Digital Glass Door Lock

    Adams Rite RITE Touch® RT2050 Digital Glass Door Lock

    Elevate interior glass door security with a sleek, surface-mounted digital lock offering touchscreen PIN, biometric fingerprint and MIFARE credential access without glass drilling.

  • GroundAware® GA1360LH 2D 360-Degree Radar

    Protect up to 750 acres with a single sensor, delivering extended 1 km+ target detection, all-weather tracking and automated intrusion deterrence for critical infrastructure.