For AI, the Worm Has Turned – How AI Companies Need to Think About Regulation Going Forward
In 2017, then-Missouri Attorney General (and now prominent US Senator) Josh Hawley announced an antitrust investigation into Google. While Missouri may not have had the power to do all that much on the antitrust front, the symbolism was clear. Politically, Google went from being everyone’s favorite to Hawley seeing that his base wanted the mega tech platforms scrutinized.
Google is still a very successful and popular company, but, to me, Hawley’s announcement represented an overall turning point towards tech: a sector that had been the recipient of lavish praise for years was now in the crosshairs. Tech founders who enjoyed endless magazine covers and lavish praise now faced tremendous scrutiny and antagonism. Things had changed.
Last week, a similar turning point occurred. President Trump announced a new federal policy that requests early government access to advanced AI models thirty days before they’re released so the government can evaluate them. The executive order establishes a new benchmarking process to determine the advanced cyber capabilities of AI models. It’s a voluntary program that’s very industry-friendly, so it’s pretty mild as far as most executive orders go. But for someone who has been a staunch proponent of frontier models and their hyperscalers like OpenAI, it’s a clear indication that Trump recognized that public sentiment has shifted so strongly against AI that he needed to do something.
In fact, a March 2026 University of Pennsylvania survey showed that just 17% of Americans think AI will have a positive impact, compared to 42% who expect it to be negative. Other polling puts the share of Americans who fear AI could eventually threaten humanity at over 75%. Nearly half of those surveyed in other polls oppose new data centers being built in their own communities. And that’s before job losses land at scale. Goldman Sachs pegs near-term direct displacement risk at around 2.5% of US employment, rising to 6–7% under broader AI adoption. Other forecasts are far more pessimistic. 41% of working Americans already say they’re worried about losing their job or having their hours cut because of AI.
New rules and regulations are hitting from every direction: states are passing everything from data center permitting and zoning limits to constraints on what chatbots can say and do to restrictions on AI in hiring and health care. For example:
New York’s legislature just passed a one year moratorium on new data center construction (legislation was also passed, but then vetoed, in Maine and Vermont).
Florida, Utah, California, Ohio, Washington and Oklahoma all enacted new regulations on energy use cost allocation from data centers.
Washington, Oregon, Maine, Idaho and Nebraska all passed bills regulating what chatbots can say and do.
Indiana and Arizona passed bills regulating AI in health care.
National Association of Insurance Commissioners (NAIC) model rules mandating that all existing insurance laws and regulation apply to AI and setting new expectations for governance, documentation, testing, and third-party oversight have been adopted in 24 states.
Note that there is virtually no partisan divide in the states above. New laws and regulations are coming fast from red states and blue states alike. Why? Because politicians respond to exactly one thing: what impacts their next election. When voters hate something this much, politicians hate it too. And if a Democrat wins the presidency in 2028 (Kalshi has it at 59-41 Democrat), it very well may come amid a rising wave of anti-AI sentiment — leading to tough new federal restrictions stacked on top of everything that states and cities are already doing.
The regulatory environment for AI today is probably the most permissive it will ever be. For AI companies ranging from hyperscalers to data center builders to applications in every conceivable sector, that means understanding the current political environment and planning accordingly. It means taking the risk of regulation and legislation very seriously and assuming that the normal political tools like campaign donations and attack ads are not going to be nearly enough to stop it.
It means having a deep understanding of what laws might pass, who is supporting them, what those politicians care about and how to develop an argument and a campaign plan to ensure that what you need is seen and reflected in their work. It also means using the current regulatory environment and negative public sentiment to put forth new regulations that impose standards that your product meets (and ideally only your product, creating a regulatory moat) and using that new regulation to help create new markets.
There’s a lot of opportunity in this new environment. And there’s a lot of risk. The smart founders and companies will see it, plan for it and benefit from it. The arrogant and clueless ones will suffer.