Federal AI Preemption Talks: OpenAI Subpoena Shows What States Could Lose

A coalition of state attorneys general opened an investigation into OpenAI on Friday, June 12, while the White House and Congress negotiate a deal that could preempt some state artificial intelligence laws. The collision makes the stakes concrete: Washington is discussing limits on state AI authority at the same moment states are using that authority to demand answers from ChatGPT's maker.

New York Attorney General Letitia James sent OpenAI a subpoena seeking documents about advertising, user engagement and retention, consumer and health data, minors and seniors, model sycophancy, and company policies, The Wall Street Journal reported. OpenAI said it takes state attorneys general's concerns seriously and intends to engage constructively.

Axios reported that the White House is negotiating federal preemption of some state AI laws in exchange for supporting congressional technology priorities, including the Kids Online Safety Act. Sen. Marsha Blackburn of Tennessee is leading the talks, according to a spokesperson cited by Axios.

No agreement, final bill text, or defined preemption scope exists publicly. State AI laws remain enforceable, and the OpenAI subpoena is an investigation, not a finding of wrongdoing.

What Does Federal AI Preemption Mean?

Preemption determines when federal law displaces state law. Congress can expressly identify the state rules it wants to override and write carve-outs preserving others. A clause targeting frontier-model development could leave employment or lending rules intact; a broader clause could constrain state investigations, chatbot-safety requirements, or consumer remedies.

The mechanism is more precise than the phrase "one national standard" suggests. Courts examine the enacted text, the federal law's purpose, and whether state requirements conflict with it. Until Congress passes a statute, executive policy preferences alone do not erase state law.

The White House's national AI framework asks Congress to preempt state AI laws it considers unduly burdensome while preserving traditional state powers involving fraud, child protection, and consumer protection. Friday's subpoena immediately tests those proposed boundaries because it covers both ordinary consumer issues and the behavior of OpenAI's models.

OpenAI Subpoena Shows Why Carve-Out Language Will Decide the Fight

The reported document demands span privacy, advertising, minors, health data, engagement, retention, and model sycophancy. Some categories fit squarely within traditional consumer-protection enforcement. Others can be characterized as oversight of how an AI model is designed, tested, and operated, which is the layer federal preemption advocates most want governed nationally.

A future statute could preserve state fraud investigations while blocking AI-specific testing mandates. It could protect state authority over companies deploying AI in employment, healthcare, lending, or education while reserving frontier-model development for federal regulators. It could also make broad assurances about consumer protection meaningless by defining any investigation of model behavior as a prohibited state AI rule.

That is why the missing bill text matters more than the reported bargain. Businesses cannot know which compliance programs would disappear, and consumers cannot know which remedies would survive, until lawmakers define the line between generally applicable law and AI-specific regulation.

Child Online Safety Has Become the Negotiating Currency

The latest talks reportedly include the Kids Online Safety Act, which would impose duties on covered platforms and require safeguards for minors. Pairing that legislation with AI preemption creates a difficult exchange: lawmakers seeking protections for children may be asked to limit state AI authority to secure them.

The pairing also demonstrates how legislative coalitions are built. Federal AI preemption has repeatedly faced resistance from state officials and lawmakers across party lines, while children's online safety commands broader support. Combining them can attract votes that a standalone preemption bill would not.

It can also obscure the tradeoff. Child-safety legislation governs how platforms treat minors; AI preemption governs which level of government may regulate a broad technology. Passing both in one package does not guarantee that the protection gained is equivalent to the state authority surrendered.

State AI Laws Regulate Different Layers, Not One Patchwork Rule

Colorado regulates high-risk AI used in consequential decisions and requires risk management, notices, and protections against algorithmic discrimination. California and New York impose transparency and safety duties on large frontier-model developers. Texas prohibits specified harmful uses and gives its attorney general enforcement authority.

These laws operate at different layers of the AI stack. Developer rules govern companies building foundation models. Deployment rules govern employers, lenders, hospitals, and schools using them. Generally applicable consumer-protection statutes can reach deceptive marketing or unsafe conduct whether or not a product contains AI.

The bipartisan Great American AI Act discussion draft released in early June proposed a three-year freeze on state laws regulating model development rather than every law touching AI. Axios reported that the new White House negotiations make that House draft unlikely to become the main vehicle for AI policy this Congress.

Businesses Still Must Follow Existing State AI Laws

Previous attempts to impose broad federal limits have failed. A proposed 10-year moratorium on state AI laws was removed from a 2025 reconciliation package after overwhelming Senate opposition. State lawmakers argued that Congress should not erase enforceable local protections without replacing them with a meaningful federal standard.

Technology companies generally support national uniformity because it reduces compliance complexity. One testing, reporting, and disclosure regime is cheaper to operate than different state definitions and deadlines. States answer that uniformity can become a weak regulatory ceiling, preventing them from responding when federal rules lag behind new harms.

For now, companies must continue complying with applicable state laws, and state attorneys general retain their investigative powers. Any eventual federal deal should be evaluated with Friday's subpoena in mind: not merely by the national rules it promises, but by whether states could still investigate advertising, data handling, child safety, and model behavior when the next AI risk appears.

Frequently Asked Questions

What is federal preemption of state AI laws?

Federal preemption means an enacted federal statute displaces some conflicting state requirements. Its effect depends on the final text, including which AI laws are covered and which consumer-protection, child-safety, and investigative powers remain.

Has Congress blocked state AI laws?

No. Negotiations are reportedly underway, but no final agreement or enacted legislation exists. Current state laws remain operative unless separately delayed, invalidated by a court, or preempted by a future federal statute.

Why are state attorneys general investigating OpenAI?

The Wall Street Journal reported that a coalition opened an investigation through a New York subpoena seeking information about advertising, data handling, minors, seniors, engagement, and model behavior. OpenAI said it intends to engage constructively.

Could federal preemption stop the OpenAI investigation?

That cannot be determined without bill text. A narrow statute could preserve generally applicable state investigations, while a broader clause could restrict oversight specifically directed at AI model design or behavior.