Hundreds of AI bills have been introduced across U.S. state legislatures since 2019. Most die in committee. Some pass one chamber and stall. A select few become law. What separates the winners from the losers?
We analyzed every state-level AI bill tracked in our Bill Tracker, cross-referenced with Brookings Institution research on AI legislation, to identify the factors that predict passage. The results are clear — and they have practical implications for anyone trying to get AI legislation enacted (or blocked).
The Headline Number: 38.6%
Bills framed around "responsible governance" pass at a rate of 38.6%. That's the single highest passage rate of any AI bill category. For context:
- Responsible governance: 38.6% passage rate
- Government AI adoption: ~28% passage rate
- Innovation & competitiveness: ~22% passage rate
- Consumer protection: ~18% passage rate
- Sector-specific regulation: ~15% passage rate
- Rights & prohibitions: ~8% passage rate
What counts as "responsible governance"? These bills typically establish study commissions, create AI advisory councils, require state agencies to inventory their AI use, or direct a state body to develop AI guidelines. They sound regulatory without actually imposing binding requirements on the private sector. That's the sweet spot for passage.
Factor 1: Framing and Language
The language used in a bill's title and summary is a surprisingly powerful predictor of its fate:
- Bills containing "study," "commission," or "task force" pass at roughly 3x the rate of bills containing "ban," "prohibit," or "moratorium"
- Bills using "transparency" pass at higher rates than those using "accountability" — even when the substantive requirements are similar
- Bills framed as "enabling" (setting up frameworks for AI use) pass at higher rates than bills framed as "restricting" (limiting AI use)
This suggests that legislative framing is doing real work. The same policy idea — say, requiring disclosure when AI is used in hiring decisions — might pass or fail depending on whether it's framed as "transparent AI hiring practices" (enabling) or "prohibition on undisclosed algorithmic hiring" (restricting).
Factor 2: Bipartisanship (or Lack Thereof)
As we detailed in our analysis of AI's partisan divide, bipartisanship is extremely rare in AI legislation. But when it exists, it's powerful:
- Bipartisan bills: ~60% passage rate (small sample size — only 3 bills — but notable)
- Single-party Democratic bills: ~20% passage rate in Democratic-controlled legislatures, near 0% in Republican-controlled ones
- Single-party Republican bills: ~25% passage rate in Republican-controlled legislatures, near 0% in Democratic-controlled ones
The takeaway is straightforward: if you want an AI bill to pass, get a co-sponsor from the other party. Almost no one is doing this.
Factor 3: State Characteristics
Some states are simply more likely to pass AI legislation than others. The state-level predictors include:
- Median household income: Higher-income states pass more AI bills, likely because they have larger tech sectors and more constituent demand for AI policy
- Tech sector employment share: States with larger tech workforces are more likely to act on AI, though the direction of action (pro-regulation vs. pro-innovation) varies
- Legislative professionalism: Full-time, well-staffed legislatures pass more AI bills than part-time citizen legislatures — they simply have more bandwidth to address emerging issues
- Political lean: Surprisingly, this is less predictive than the other factors. Both red and blue states pass AI bills; they just pass different kinds
Use our State Comparison Tool to see how these characteristics map onto actual legislative outcomes.
Factor 4: Topic and Scope
The specific topic of an AI bill matters enormously:
- Highest passage rates: Government AI use/procurement, AI literacy/education, study commissions
- Medium passage rates: Deepfake disclosure, AI in elections, healthcare AI
- Lowest passage rates: General-purpose AI regulation, algorithmic accountability mandates, AI moratoriums
Narrow, sector-specific bills outperform broad, horizontal bills. A bill requiring AI disclosure in political advertising is far more likely to pass than a bill establishing comprehensive AI accountability standards across all sectors. Legislators are more comfortable regulating AI in a specific context they understand than regulating AI in general.
Building a Predictor
We combined these factors into a simple predictive model — available on our Bill Predictor tool. Input a bill's characteristics (state, sponsor party, framing, topic, bipartisanship) and the model estimates passage probability.
The model isn't perfect — politics is inherently unpredictable. But it captures the structural factors that tilt the odds. A bipartisan, narrowly-scoped, responsible-governance-framed bill in a high-income state with a professional legislature has roughly a 50-60% chance of passage. A single-party, broad-scope, rights-and-prohibitions bill in a part-time legislature has less than a 5% chance.
Implications
For advocates who want AI regulation: the data says go narrow, go bipartisan, go governance. Grand comprehensive bills make great press releases and terrible laws (because they never pass). Incremental, focused bills actually move.
For industry players who want to prevent regulation: the data says your biggest risk isn't the sweeping AI accountability acts (those die on their own). It's the quiet, narrow, bipartisan bills that create study commissions — because those commissions often recommend binding rules in their second year.
Track every bill's status and predicted probability on our Bill Tracker.