U.S. AI Regulatory Patchwork: Preemption Showdowns, State-Level ADMT Mandates, and the Federal Preemption Carve-Out

U.S. AI Regulatory Patchwork: Preemption Showdowns, State-Level ADMT Mandates, and the Federal Preemption Carve-Out

As the United States enters the latter half of 2026, the absence of a cohesive federal AI framework has catalyzed a highly complex state-level regulatory patchwork¹. Far from simplifying compliance, recent legislative overhauls and pending bills have created conflicting timelines and operational requirements that enterprise legal and risk teams must navigate immediately.

A critical development in late 2026 is the emergence of a "preemption showdown" between federal deregulation efforts and state-level employment protections. While the federal government is actively pushing to preempt state-level AI restrictions, the most comprehensive federal legislative proposal explicitly carves out state authority over employment-related AI, ensuring that the state-level patchwork will remain a dominant compliance headache for employers.

1. State-Level Legislative Overhauls and Mass Layoff Disclosures

• Connecticut SB 5 (Public Act No. 26-15): Signed into law on May 27, 2026, Connecticut’s workplace AI law moves on an aggressive timeline. Starting October 1, 2026, covered employers must disclose whether artificial intelligence or automated decision-making technology contributed to certain mass layoffs or reductions in force. This creates an immediate compliance trigger for any enterprise planning a RIF in Connecticut.
• Colorado SB 26-189: Signed on May 14, 2026, this bill completely repealed and replaced the landmark Colorado Artificial Intelligence Act (CAIA / SB 24-205). Originally slated to take effect in 2026, the overhaul delayed the effective date to January 1, 2027. SB 26-189 removes the original law’s onerous requirements for structured risk management programs and annual algorithmic impact assessments. Instead, it implements a narrower, procedural framework focusing on notice and disclosure. This legislative reset was heavily influenced by xAI v. Weiser (filed April 9, 2026), a federal lawsuit by Elon Musk’s xAI seeking to enjoin the CAIA on First Amendment and Commerce Clause grounds, which has blunted the state's enforcement appetite.
• California "No Robo Bosses Act" (SB 947): Passed by the California Legislature in late August 2026, this bill is currently awaiting action by Governor Gavin Newsom, who has until September 30, 2026, to sign or veto it. Reintroduced after a 2025 veto of a broader version, SB 947 would prohibit employers from relying solely on automated systems to fire, discipline, or manage employees, requiring a mandatory human-in-the-loop oversight mechanism.
• California CPPA ADMT Regulations: Finalized by the California Privacy Protection Agency, these regulations governing businesses’ use of Automated Decision-Making Technology (ADMT) in significant decisions (including hiring, termination, and compensation) are set to become legally enforceable on January 1, 2027, establishing broad notice, access, and opt-out rights for California workers.

2. The Federal Counterweights: Preemption vs. State Carve-Outs

The federal government has attempted to blunt this state-level patchwork through administrative and legislative channels:

• Executive Order 14365: Signed by President Trump on December 11, 2025, this EO identified "excessive state regulation" as a threat to American AI leadership and directed federal agencies to maximize voluntary, innovation-friendly frameworks.
• The Great American Artificial Intelligence Act of 2026: On June 4, 2026, Representatives Jay Obernolte (R-CA) and Lori Trahan (D-MA) released a 269-page bipartisan discussion draft of this sweeping federal AI bill. Designed to establish a unified national policy, the draft includes robust federal preemption provisions to override state AI laws. Crucially, however, the draft carves out state authority over employment-related AI use from its preemption provisions.

As analyzed by Epstein Becker Green in September 2026:

"The discussion draft of the most comprehensive federal AI bill in Congress would carve out state authority over employment-related AI use from its preemption provisions. As a result, existing state AI laws governing employment are highly likely to remain intact even if this legislation is enacted as drafted."

Compliance Action Items for Enterprise Risk Teams

To manage this rapidly shifting landscape, enterprise risk and legal departments should adopt a multi-jurisdictional compliance framework:

1. AI Tool Inventory: Document every predictive or generative tool used in hiring, performance management, compensation, and termination.
2. RIF Disclosure Readiness: Establish protocols to identify and document whether automated tools or productivity metrics (such as those at issue in Meta's Landmark AI Layoff Lawsuit: Disparate Impact, Leave-Aware Design, and the "Serious Questions" Standard) played any role in RIF selections, ensuring compliance with Connecticut's October 1, 2026 disclosure deadline.
3. ADMT Opt-Out and Notice Protocols: Prepare notice, access, and opt-out workflows for California and Colorado workers ahead of the January 1, 2027 effective dates.
4. Human-in-the-Loop Checkpoints: Ensure all critical employment decisions (disciplinary actions, terminations) have documented human reviews to mitigate "No Robo Bosses" exposure and rebut automated disparate impact claims.