AI Frontier Institute

ARTIFICIAL INTELLIGENCE SAFETY

View primary source →

Illinois SB3261, introduced February 3, 2026 by Sen. Mary Edly-Allen, creates the Artificial Intelligence Public Safety and Child Protection Transparency Act (Section 1), effective January 1, 2027 (Section 99). The bill targets two categories of regulated entities: 'large frontier developers'—companies with at least $500,000,000 in annual revenue that have trained a foundation model using more than 10^26 computing operations—and 'large chatbot providers'—companies with at least $25,000,000 in annual revenue that operate a covered chatbot with at least 1,000,000 monthly active users foreseeably accessible to minors (Section 10). Under Section 15, both must write, implement, and publicly post a 'public safety and child protection plan' detailing how they assess and mitigate 'covered risks,' which include catastrophic risks (e.g., assisting in creation of weapons of mass destruction, autonomous cyberattacks, or loss of model control causing death or $1 billion in damages) and child safety risks (behavior toward minors that could cause death, bodily injury, or severe emotional distress). Large frontier developers must also publish summaries of catastrophic risk assessments before deploying new or substantially modified frontier models, and may not deploy a frontier model that poses unreasonable catastrophic risk. Under Section 20, frontier developers must report critical safety incidents to the Attorney General within 15 days, or within 24 hours if an imminent risk of death or serious injury exists; large chatbot providers must report child safety incidents within 15 days; and large frontier developers must submit quarterly summaries of internal catastrophic risk assessments. Section 30 protects whistleblowers, requiring large frontier developers to maintain anonymous internal disclosure processes with monthly status updates. Section 35 mandates annual third-party audits of large frontier developers assessing plan compliance, clarity, and the reasonableness of redactions, with audit reports retained for five years and available to the Attorney General. Civil penalties under Section 40 reach up to $1,000,000 per violation for large frontier developers and $50,000 per violation for large chatbot providers, enforced by the Attorney General. Section 25 directs the Attorney General to review and update key definitions annually beginning January 1, 2028. Section 50 allows the Attorney General to establish alternative compliance pathways for entities that comply with substantially equivalent federal or other-state law.

← Back to the tracker