AI product management · Industry brief

Businesses operating nationally face immediate compliance exposure under a patchwork of state AI laws that took effect or are taking effect in 2026, despite the absence of comprehensive federal AI regulation. Texas's TRAIGA (effective January 1, 2026) prohibits intentional deployment of AI for behavioral manipulation and deceptive deepfakes, enforced by the state attorney general; Illinois's employment-AI law HB 3773 requires notice when AI is used in employment decisions and prohibits discriminatory effects; California's SB 53 imposes safety and transparency frameworks on "frontier developers" training models above specified compute thresholds; and Connecticut's CART Act (effective October 1, 2026) imposes staggered obligations through January 2028 on specified high-risk uses including automated employment decision technology and AI chatbots. Colorado's revised SB 26-189 (effective January 1, 2027) expands coverage of Automated Decision-Making Technology and voids contract terms shielding developers from liability for discriminatory outcomes. A December 2025 White House executive order directing DOJ challenges to state AI laws on constitutional grounds has not prevented enforcement. Mid-Atlantic and other multistate companies are in scope based on customer and employee location rather than company headquarters location.

On September 29, 2026, the White House announced a voluntary framework called "The White House Accord on Super Intelligence" with leading AI developers, committing signatories to adopt internal controls for frontier AI systems, engage independent external auditors, and establish independent board-level oversight. The Accord, though not legally binding, represents a significant policy shift from the Administration's earlier skepticism of third-party AI audits and is expected to operate as a market standard, influencing vendor contracts, board governance structures, and audit practices across the industry. The framework's four-layer control structure and potential codification into future legislation signals that the Accord may inform regulatory developments, with the FTC reportedly initiating a broad investigation into frontier AI developers the day after the Accord was signed, and prospective auditors already in discussions with the Department of Commerce regarding formalized government-audit provider relationships (source: Alston & Bird LLP privacy advisory).

California enacted Senate Bill 574, signed by Governor Newsom on September 30, 2026, establishing the first state rules explicitly regulating attorneys' use of generative AI, effective January 1, 2027. The law amends Business and Professions Code Section 6068.1 and Code of Civil Procedure Section 128.7 to impose several compliance requirements: attorneys cannot delegate the practice of law to AI; they must restrict confidential client information from being entered into AI systems unless access is limited to authorized persons bound by confidentiality obligations; they must verify the accuracy of AI outputs, including all case and statutory citations; and they must personally verify every citation in California court filings. Additionally, Section 128.7 violations are subject to sanctions including penalties and attorney's fees. Attorneys must also disclose their use of generative AI to courts for documents submitted, though the statute leaves the form and scope of disclosure unresolved pending interpretation.