Bill > SB833

This bill requires state agencies responsible for critical infrastructure to establish rigorous human oversight mechanisms for artificial intelligence (AI) systems by July 1, 2026. Specifically, the bill mandates that human oversight personnel must monitor AI systems in real-time and review and approve any actions proposed by the system before execution, with some exceptions for systems where immediate intervention could cause disruption. The Department of Technology must develop specialized training in AI safety protocols for oversight personnel, who will be required to conduct annual assessments of AI systems, evaluating their performance, safety, and potential risks, including scenarios that could lead to mass casualty events. The bill also introduces a comprehensive AI adverse event reporting system, requiring entities to report incidents within specified timeframes depending on the severity, such as within four hours for urgent threats to public health or safety, and within 24 hours for events involving death or serious injury. Entities that fail to report such events may face civil penalties up to $500 for each seven-day period of non-compliance. The legislation defines key terms like "critical infrastructure" (which includes sectors such as transportation, energy, communications, and emergency services) and "AI adverse events" (incidents causing death, serious injury, significant infrastructure disruption, or substantial financial loss). The bill aims to ensure responsible AI deployment, enhance public safety, and create a framework for transparent and accountable use of AI in critical state infrastructure.

Budget and Finance, Business and Industry, Government Affairs

https://leginfo.legislature.ca.gov/faces/billStatusClient.xhtml?bill_id=202520260SB833