Governor Gavin Newsom signed an executive order this week telling California state agencies to independently review federal AI supply-chain risk designations—a direct response to the Trump administration labeling San Francisco-based Anthropic a supply-chain risk.
Meanwhile, Tennessee became the first state to outlaw AI systems that impersonate mental health professionals. South Carolina’s House passed its anti-addictive social media bill 114-0. And Maryland moved forward on banning AI-driven surveillance pricing.
The White House wants one national AI standard. The states are writing their own.
California: Newsom Fires Back
Newsom’s executive order is a pointed rebuttal to the federal framework released on March 20. After the Department of Defense designated Anthropic as a supply-chain risk, Newsom ordered California to conduct its own independent review before restricting any company from state contracts.
What the order requires:
- State agencies must develop AI contract standards addressing CSAM generation, civil liberties violations, and bias
- State employees get access to approved generative AI tools
- California’s Digital Strategy must be updated to integrate AI across government services
- New guidance on watermarking AI-generated imagery and video
Newsom’s office stated the obvious contrast: “Unlike the Trump administration, California remains committed to ensuring that AI solutions adopted…cannot be misused by bad actors.”
The politics are layered. Union leaders want stronger worker protections from AI before backing Newsom’s presidential ambitions. Tech donors are pouring money into California politics ahead of midterms. Newsom is threading a needle—pro-innovation, pro-regulation, anti-federal-overreach—all at once.
Over 20 state departments are already developing or using Poppy, a generative AI assistant for state employees. California isn’t just regulating AI. It’s deploying it.
Tennessee: AI Can’t Play Therapist
Governor Bill Lee signed SB 1580 on April 1, making Tennessee the first state to explicitly ban AI systems from representing themselves as qualified mental health professionals. The bill passed the Senate 32-0 and the House 94-0. No opposition.
The law takes effect July 1 and carries real teeth: violations trigger the Tennessee Consumer Protection Act, with penalties of $5,000 per violation plus injunctive relief and private right of action. Users can sue.
This matters because the AI therapy market has exploded. Apps like Woebot, Wysa, and dozens of ChatGPT wrappers market themselves as mental health support tools. Some stop short of claiming to be licensed professionals. Others don’t. Tennessee drew a bright line: if your AI system advertises itself as a qualified mental health professional, you’re breaking the law.
Expect other states to follow. Mental health AI regulation is one of the few areas with genuine bipartisan consensus—the unanimous votes prove it.
The Bills That Moved This Week
The Transparency Coalition’s April 3 tracker documents a surge of activity as state legislatures approach session deadlines.
Signed into Law
| Bill | State | What It Does |
|---|---|---|
| SB 1580 | Tennessee | Bans AI impersonation of mental health professionals |
| HB 727 | Idaho | Adds synthetic media to video voyeurism laws |
Passed Chambers, Heading to Governor
| Bill | State | What It Does |
|---|---|---|
| SB 1227 | Idaho | Generative AI rules for public education (passed 26-8, 62-6) |
| SB 1297 | Idaho | Conversational AI Safety Act (chatbot regulation) |
| HB 542 | Idaho | Anti-addictive social media protections |
| SB 540 | Georgia | Chatbot disclosure and child safety requirements |
| SB 444 | Georgia | Bans insurance decisions based solely on AI |
Passed One Chamber
| Bill | State | What It Does |
|---|---|---|
| HB 4591 | South Carolina | Anti-addictive social media (House: 114-0) |
| HB 883 | Maryland | Bans AI chatbot therapy (House: 110-23) |
| SB 8 | Maryland | Deepfake protections (Senate: 45-0) |
| SB 300 | California | Strengthens chatbot laws (Senate: 38-0) |
| HB 1782 | Hawaii | Chatbot safeguards for minors |
| SB 3001 | Hawaii | AI disclosure requirements |
Advancing in Committee
| Bill | State | What It Does |
|---|---|---|
| HB 148 | Maryland | Bans AI-driven surveillance pricing and wage setting |
| HB 1263 | Colorado | Chatbot safety for minors |
| HB 1210 | Colorado | Restricts intimate data for financial inferences |
| HB 2311 | Arizona | Kids chatbot safety |
| SB 1786 | Arizona | AI-generated media provenance requirements |
| AB 1898 | California | Workplace AI surveillance notice |
| SB 867 | California | Bans companion chatbots in children’s toys |
| LB 1185 | Nebraska | Chatbot safety (attached to Agricultural Data Privacy Act) |
| SB 760 | Michigan | Kids chatbot safety |
That’s not a trickle. That’s a flood.
Surveillance Pricing: Maryland Takes Aim
Maryland’s HB 148 deserves special attention. The bill would ban businesses from using personal surveillance data with automated systems to set individualized prices for consumers or wages for employees.
In plain terms: if a company uses your browsing history, location data, or behavioral profile to charge you a different price than the person standing next to you, that would be illegal in Maryland.
The bill passed out of committee 9-0 on April 2. Key exemptions include cost-based pricing differences and voluntary discount programs where the consumer knowingly shares data. Employers can still set wages based on job-specific data, but must disclose what data the automated system uses and how.
Maryland isn’t alone. New Jersey introduced similar legislation in March. The concept of “surveillance pricing”—algorithms that set prices based on what they think you’ll pay—is drawing bipartisan scrutiny.
The Federal Standoff Deepens
The Trump administration’s December 2025 executive order established an AI Litigation Task Force within the DOJ to challenge state AI laws in federal court. The March 2026 National AI Policy Framework recommended Congress preempt state laws across seven policy pillars.
But the framework’s own carve-outs for child protection, consumer protection, and state procurement cover most of what states are actually passing. The chatbot bills, deepfake protections, and AI therapy bans all arguably fit within the exempted categories.
The real test will come if the DOJ task force actually files suits. So far, it hasn’t. And Congress hasn’t passed preemption legislation either—declining to include it in both the One Big Beautiful Bill Act and the NDAA.
Meanwhile, Newsom’s executive order is an explicit dare: California will make its own AI risk assessments regardless of what the federal government says.
The Emerging Pattern
Five themes dominate state AI legislation this spring:
- Chatbot safety for minors — Disclosure requirements, parental consent, self-harm protocols. Oregon set the template; everyone’s copying it.
- AI therapy restrictions — Tennessee drew first blood. Maryland’s chatbot therapy ban passed 110-23. Others are moving.
- Deepfake and synthetic media — Provenance data, watermarking, video voyeurism updates. Near-universal support.
- Surveillance pricing — Maryland and New Jersey leading. Could be the next wave.
- State vs. federal authority — California vs. the White House. States aren’t blinking.
Session deadlines create urgency. Georgia’s session ends April 6. Idaho’s ends April 10. Nebraska’s ends April 17. Bills that don’t pass in the next two weeks die.
What to Watch
The DOJ’s AI Litigation Task Force has been quiet. If it starts filing suits against specific state laws, the preemption fight moves from policy papers to courtrooms. That would be a different game entirely.
California’s legislature has dozens of AI bills in committee, with hearings scheduled through April. SB 867 (companion chatbot ban in children’s toys) has a committee hearing on April 6. AB 1898 (workplace AI surveillance notice) already cleared committee.
And the surveillance pricing movement is just starting. If Maryland passes HB 148, expect a wave of similar bills in 2027 legislative sessions.
The federal government publishes frameworks. The states pass laws. Four days into April, that pattern hasn’t changed.