TL;DR: On December 20, 2025, New York signed the RAISE Act requiring big AI companies to publish safety protocols and report incidents within 72 hours. Nine days earlier, Trump signed an executive order creating an "AI Litigation Task Force" to sue states that regulate AI. The battle lines are drawn: states want accountability, the federal government wants to block it, and AI companies are watching to see who wins.
The Timeline
December 11, 2025: Trump signs Executive Order "Ensuring a National Policy Framework for Artificial Intelligence" [1].
December 20, 2025: New York Governor Kathy Hochul signs the RAISE Act (Responsible AI Safety and Education Act) [2].
Nine days apart. Two completely opposed visions of AI governance. One legal collision course.
What the RAISE Act Does
Who It Covers
The law applies to AI developers who have spent more than $100 million in computational resources to train advanced AI models [3]. That means companies like:
- OpenAI
- Google/DeepMind
- Anthropic
- Meta AI
- Other "frontier" AI developers
Key Requirements
Safety Protocols: Large AI developers must create and publish safety plans covering severe risks, including:
- Use of AI to create biological weapons
- AI-enabled criminal activity
- Harms that could cause death or injury to 100+ people
- Damages exceeding $1 billion
Incident Reporting: Companies must report critical safety incidents to the state within 72 hours—a much tighter window than California's 15-day requirement [4].
Transparency: Safety plans must explain in detail how developers will handle (not just "approach") various risks.
Enforcement
- First violation: Up to $1 million fine
- Subsequent violations: Up to $3 million per violation
- Attorney General can bring civil actions for failure to report or false statements
New Oversight Office
The law creates a new office within the Department of Financial Services to monitor AI development and issue annual reports assessing large AI developers [2].
What Got Watered Down
The original bill had teeth. Tech industry lobbying filed them down [5]:
- Original fines: $10 million first violation, $30 million subsequent
- Final fines: $1 million first violation, $3 million subsequent
Lawmakers passed the original bill in June. Hochul proposed weakening it. Eventually, she signed the original version while lawmakers agreed to make her requested changes next year. The companies got what they wanted: a promise of future weakening.
What Trump's Executive Order Does
The AI Litigation Task Force
The order directs the Department of Justice to create an "AI Litigation Task Force" with one job: sue states over their AI regulations [6].
The task force will challenge state laws the Attorney General deems "unlawful" or inconsistent with the federal government's "minimally burdensome" AI policy.
Targeting Specific Laws
The executive order explicitly names [7]:
- Colorado's AI Act: Called out as an example of "problematic" state regulation. The law addresses algorithmic discrimination and was set to take effect February 2026 (now delayed to June 2026).
- California's AI laws: Data transparency and reporting requirements for generative AI systems.
Additional Weapons
Beyond lawsuits, the order deploys multiple pressure tactics:
- FTC and FCC pressure: Directed to work with DOJ to circumvent "onerous" state and local regulations
- Funding threats: Commerce Secretary must study whether federal rural broadband funding can be withheld from states with unfavorable AI policies
- Administrative reinterpretation: Using existing federal laws to override state authority
David Sacks' Role
Trump's AI advisor David Sacks—a venture capitalist with investments across the AI industry—will recommend which state laws warrant exemption and direct efforts to work with Congress on national AI legislation [1].
The person deciding which regulations survive has financial interests in the companies being regulated.
Will the Executive Order Work?
Legal Experts Say: Probably Not
Constitutional scholars and tech policy researchers are skeptical [1]:
The preemption problem: For a state law to be preempted by federal law, Congress generally must have legislated in that space. Congress has not enacted comprehensive AI legislation. An executive order isn't a law.
Republican governors agree: Florida's Ron DeSantis noted that "an executive order doesn't/can't preempt state legislative action" [1].
Supreme Court precedent: A 2023 decision supported California's authority to regulate industries affecting other states, suggesting courts may side with state regulatory power [6].
What States Can Do
States will likely challenge the executive order on federalism and 10th Amendment grounds. States traditionally regulate:
- Employment
- Civil rights
- Consumer protection
All areas AI regulation touches. California and New York have already signaled they'll fight federal overreach [6].
The Current Legal Reality
All existing state AI laws remain enforceable unless and until [6]:
- A court blocks them through an injunction, OR
- Congress passes a federal law that explicitly preempts them
Neither has happened. The executive order is a threat, not a legal ruling.
Who's On Which Side
Supporting State Regulation
- New York (RAISE Act signed)
- California (multiple AI laws enacted)
- Colorado (AI Act addressing algorithmic discrimination)
- Consumer protection advocates
- Civil rights organizations
- Child safety groups (though frustrated by weak exemptions)
Supporting Federal Preemption
- Trump administration
- OpenAI (supports "federal legislation" while opposing state rules)
- Anthropic (similar position)
- Tech industry lobbyists
- Venture capital investors in AI
The Irony
OpenAI and Anthropic publicly expressed support for New York's RAISE Act while also calling for federal legislation [3]. They're playing both sides: appear to support safety while lobbying for federal preemption that would eliminate stronger state laws.
Child Safety Concerns
The executive order claims to exempt child safety protections from preemption. But child safety advocates aren't convinced.
Michael Toscano of the Institute for Family Studies called the order "a huge lost opportunity" [1]. Conservative organizations focused on child safety joined liberals in criticizing the approach.
Why This Matters
The Race to the Bottom
If the federal government successfully blocks state AI regulation:
- No state can require safety reporting
- No state can address algorithmic discrimination
- No state can mandate transparency
- AI companies face no accountability until Congress acts (which may be never)
The Surveillance Connection
AI powers the surveillance state. Facial recognition, predictive policing, immigration enforcement—all increasingly run on AI. State regulations that require transparency and accountability for AI directly affect surveillance systems.
Colorado's AI Act addresses algorithmic discrimination. That includes AI systems that discriminate in housing, employment, and law enforcement contexts. Blocking that law protects discriminatory algorithms.
The Pattern
We've seen this before:
- States try to protect residents
- Industry lobbies the federal government
- Federal government tries to override state protections
- Courts decide who wins
It happened with privacy (California's CCPA). It happened with facial recognition (Illinois's BIPA). Now it's happening with AI.
What's Coming
Legal Challenges
The executive order "is almost certain to be challenged in court" [1]. States like California and New York will likely file suits arguing:
- Executive orders can't preempt state law
- Only Congress can legislate in this space
- States retain traditional regulatory authority
Colorado's AI Act Delayed
Colorado's AI Act was supposed to take effect February 1, 2026. It's now delayed to June 30, 2026 [6]—likely giving the federal government time to mount legal challenges.
Congressional Action?
The Trump administration wants Congress to pass a federal AI law that explicitly preempts state regulation. Whether Congress acts—and what that law would look like—remains unclear.
What You Can Do
Immediate Actions
• Contact your state representatives to support strong AI regulation
• Comment on federal rulemaking processes about AI
• Support organizations fighting for AI accountability
• Track which AI companies are lobbying against state laws
Long-term Advocacy
• Support state attorneys general who challenge federal preemption
• Push for federal AI legislation with strong protections (not industry-written)
• Demand transparency from AI companies about their lobbying
• Vote for candidates who support AI accountability
The Bottom Line
New York passed a law requiring AI safety protocols. Trump signed an order to sue states for regulating AI. The legal fight is coming.
Legal experts think the executive order probably won't hold up in court. But "probably" isn't "definitely." And while the courts decide, the chilling effect is real. States may hesitate to pass new AI regulations knowing the federal government will challenge them.
That's the goal. Not to win every lawsuit—but to scare states into doing nothing.
The question is whether states like New York and California will fight anyway. So far, they're fighting.
References
- NPR - Trump is trying to preempt state AI laws via an executive order. It may not be legal (December 11, 2025)
- Governor Hochul - Signs Nation-Leading AI Legislation (December 20, 2025)
- TechCrunch - New York Governor signs RAISE Act (December 20, 2025)
- Axios - NY Gov. Kathy Hochul signs sweeping AI safety bill (December 19, 2025)
- City & State NY - Hochul signs watered down AI regs (December 2025)
- Gibson Dunn - Trump's Executive Order Seeks to Preempt State Laws (December 2025)
- Economic Policy Institute - Executive Order to challenge state AI laws (2025)