If federal action in the US isn't forthcoming on AI, state Attorneys General (AGs) may end up being a central force on regulation of AI. I tasked Opus 5.5 with helping me collect some data on the actions that the various state AGs have taken on AI, to try to understand how much appetite there is among the states to take on AI companies.
Summary
Every state office has sent some kind of signal that they are willing to take on AI companies (e.g. signing a letter to an AI company)
A smaller group of states (15) have taken at least 1 step that indicates willingness to take initiative or lead actions against AI companies (e.g. opening its own investigation)
Several states in the second group (4) may now be in flux with a new Attorney General coming into office after the 2026 midterms
Only two states have an AG who has led or acted alone on AI and has a 1 GW or more data center within the state
The state of state AGs
With the help of Opus 5.5, I collected data from the websites of state AGs on the various actions that each state has been involved in. I have Opus classify the actions each state is involved in and use these assignments to place the states into tiers 1-4, with higher tiers indicating that a state has taken more aggressive action to regulate AI. I think the move from 2 to 3 is particularly interesting, since this reflects the difference between a state passively joining actions by other AGs and an office initiating or leading actions itself. You can see the results of coding the states in this way below:
as of Nov 10, 2026
Most states have at least joined a brief, investigation or suit, although for many states the only action putting them in this category is the Character.AI bankruptcy. Some state-specific observations:
NY is only at a 2 despite having a dedicated tech unit in the office and having some state-level appetite to regulate AI with the RAISE Act.
NM is at a 1 despite the AG there having just won a large jury verdict against Meta over data privacy violations.
TX is at a 4, and has a track record of taking tech cases through multiple appeals as well as several large settlements with Meta and Google. However, Texas will have a new AG coming in after the midterms, and the state's courts may also be in flux with changes on its 15th Court of Appeals, which hears appeals from its new business court.
With TX being something of a question mark and now having an incoming Democratic AG, FL may be the state to take up leadership on tech issues among Republican states. The FL AG is currently pursuing a high-profile investigation/lawsuit against OpenAI, suggesting a strong appetite for action on AI.
Data Centers
In my data, only 2 of the tier 3/4 states also have a 1GW+ data center located in the state (TX and KY). Having a data center in the state could create economic incentives that make an AG more hesitant to oppose AI companies (which is potentially consistent with many of the data center states joining instead of leading). On the other hand, I could imagine having a data center in the state being an important jurisdictional hook, particularly in scenarios where AI capabilities accelerate rapidly and states become interested in getting a lot more aggressive on regulation. Consider a situation like the one in AI2040, where accelerating AI capabilities are rapidly changing the world and AI is a central political issue. In the absence of federal action, a coalition of states might try to implement a local version of plan S or A (i.e. stopping or significantly slowing AI development using levers within those states). AI companies seeking to continue developing ever more advanced models might attempt to get around these limitations by leaving the jurisdiction of those states. Having an AI data center located in-state would be one of the strongest limiting factors for this kind of scenario, since an AI company that is physically training or running their models on a data center in a state would almost certainly be subject to that state's jurisdiction.
Limitations
I may have coded some actions incorrectly or missed some actions that an AG has been involved in. Please feel free to comment below or reach out to me with corrections or actions I may have missed.
If federal action in the US isn't forthcoming on AI, state Attorneys General (AGs) may end up being a central force on regulation of AI. I tasked Opus 5.5 with helping me collect some data on the actions that the various state AGs have taken on AI, to try to understand how much appetite there is among the states to take on AI companies.
Summary
The state of state AGs
With the help of Opus 5.5, I collected data from the websites of state AGs on the various actions that each state has been involved in. I have Opus classify the actions each state is involved in and use these assignments to place the states into tiers 1-4, with higher tiers indicating that a state has taken more aggressive action to regulate AI. I think the move from 2 to 3 is particularly interesting, since this reflects the difference between a state passively joining actions by other AGs and an office initiating or leading actions itself. You can see the results of coding the states in this way below:
as of Nov 10, 2026
Most states have at least joined a brief, investigation or suit, although for many states the only action putting them in this category is the Character.AI bankruptcy. Some state-specific observations:
Data Centers
In my data, only 2 of the tier 3/4 states also have a 1GW+ data center located in the state (TX and KY). Having a data center in the state could create economic incentives that make an AG more hesitant to oppose AI companies (which is potentially consistent with many of the data center states joining instead of leading). On the other hand, I could imagine having a data center in the state being an important jurisdictional hook, particularly in scenarios where AI capabilities accelerate rapidly and states become interested in getting a lot more aggressive on regulation. Consider a situation like the one in AI2040, where accelerating AI capabilities are rapidly changing the world and AI is a central political issue. In the absence of federal action, a coalition of states might try to implement a local version of plan S or A (i.e. stopping or significantly slowing AI development using levers within those states). AI companies seeking to continue developing ever more advanced models might attempt to get around these limitations by leaving the jurisdiction of those states. Having an AI data center located in-state would be one of the strongest limiting factors for this kind of scenario, since an AI company that is physically training or running their models on a data center in a state would almost certainly be subject to that state's jurisdiction.
Limitations
I may have coded some actions incorrectly or missed some actions that an AG has been involved in. Please feel free to comment below or reach out to me with corrections or actions I may have missed.