California AI laws and federal preemption talks

California has enacted two rounds of AI legislation while Congress considers multiple competing federal frameworks that would preempt state laws. California signed SB 53 on September 29, 2025, creating what Brookings described as the first enforceable U.S. regulatory framework for frontier AI systems. A second package signed September 30, 2026 includes AB 1883, which bans employers from using AI to read employee emotions, bars AI-only disciplinary or termination decisions, and requires disclosure when AI drives mass layoffs; companion provisions protect physicians' clinical judgment, ban removing AI-generated watermarks, strengthen deepfake protections, and require gene synthesis companies to screen orders.

At the federal level, at least three preemption vehicles are active. H.R.5388, introduced September 16, 2025, would impose a temporary moratorium preempting state AI laws for systems in interstate commerce. The FRONTIER Act (H.R.9925), introduced July 23, 2026 by Reps. Jay Obernolte and Lori Trahan on a bipartisan basis, would require large frontier developers to undergo annual third-party compliance audits. Senator Marsha Blackburn's TRUMP AMERICA AI Act discussion draft, released March 18, 2026, would replace the state patchwork with a single federal rulebook focused on children, creators, conservatives, and communities, and would codify President Trump's December 2025 executive order on AI. The White House released a National Policy Framework on March 20, 2026 urging Congress to act, though Morgan Lewis noted it imposes no immediate compliance obligations.

Democrats are preparing rival proposals. Reps. Don Beyer and Sara Jacobs have outlined a bill to create a standard-setting federal agency with an emergency shutdown mechanism for AI models; Sen. Maria Cantwell published a framework calling for constant oversight by government and independent experts without preemption; Sen. Ruben Gallego proposed a jobs program to address AI-driven unemployment; and Sens. Bernie Sanders and Rep. Greg Casar introduced the Ban Artificial Superintelligence Act. Transformer News reported that Democrats fear Republicans may use a lame-duck session to advance the FRONTIER Act or a Thune-Klobuchar bill.

The Guardrails Alliance, a super PAC, announced $1.2 million in spending to oppose three congressional candidates backed by the pro-AI-industry super PAC Leading the Future, targeting Aaron Flint (MT-1), Jay Feely (AZ-1), and Rep. Jimmy Gomez (CA-34). A coalition of nearly 40 labor, progressive, faith, and AI safety groups led by Guardrails Action sent Congress a letter demanding enforceable government oversight not designed or controlled by major tech firms, warning of active, mobilized opposition to bills falling short of their standards. A CSAIP poll of 2,498 Americans found 61% consider voluntary AI safety commitments by companies like Anthropic and OpenAI insufficient, and 54% favor government-set and enforced rules.

As of March 2026, lawmakers in 45 states had introduced 1,561 AI-related bills, exceeding 2024's total volume, and multiple state laws now have 2026 effective dates. Anthropic publicly endorsed SB 53; OpenAI and Google did not oppose it but called for federal preemption. Dean Ball noted SB 53 covers internal lab deployments of frontier models, not only publicly released ones, and that at least one AI developer is sharing its compliance report publicly as well as with the California government.