Viewpoints

What a Democratic Congress means for AI: more scrutiny, few new laws, the states still in charge

Richard White · · 5 min read

Democrats will take both chambers in January, but the President keeps the veto through January 2029. Expect hearings, document requests and pressure on data centers rather than a federal AI law. For firms putting AI into production, the practical rulebook stays with the states, and the evidence you can produce matters more than ever.

What happened

Democrats won control of both the House and the Senate in the 3 November midterms, NBC News projected, flipping at least seven Senate seats and more than two dozen House districts, with some races still being counted. The new Congress convenes in January 2027. President Trump keeps the veto, and the Senate filibuster still requires 60 votes for most legislation.

AI and data centers were on the ballot in a way they haven’t been before. Reuters described Democratic wins in the Texas and Georgia governor’s races, where candidates campaigned partly on pausing data center development, and quoted one investor calling the vote “more like a referendum on AI than a referendum on Trump.” Texas accounts for about a quarter of US data center capacity due online by 2030, according to an estimate Reuters cited. Tech shares fell sharply on 5 November, and Meta’s president said building in the US “is going to be harder.”

The National Law Review’s reading of what comes next is the clearest I’ve seen. Oversight comes first: hearings, document requests and executive testimony on safety incidents, testing, competition and data practices from early 2027. House Democratic committee leaders have already asked five frontier AI companies to report misalignment and safety incidents to Congress within 48 hours. Bills in circulation include Sen. Maria Cantwell’s framework for NIST safety standards, pre-release audits and incident reporting, and Rep. Lori Trahan’s draft CLAIM Act, which would create a federal cause of action against developers whose AI systems cause harm. Tax writers have named AI and data center taxation, including bonus depreciation, as priorities. Kids’ online safety and chip export controls have the best odds of passing.

Democrats aren’t united on data centers. Rep. Alexandria Ocasio-Cortez and Sen. Bernie Sanders have proposed a national moratorium on large AI data centers. Others in the caucus oppose it or haven’t taken a position, and House leader Hakeem Jeffries has said a moratorium is “not a position that I’ve articulated.”

My view

The headline will be “Democrats crack down on AI.” The operating reality for most firms will be quieter than that, and more demanding in a different way.

Little new federal law, a lot of new scrutiny

A divided government rarely passes sweeping technology law, and this one starts with a President who has opposed binding AI safety mandates. Supporters of stronger rules will point to the incident reporting request and the liability draft as the start of a real framework. Skeptics will say that rules written in response to an election, rather than to evidence, risk slowing useful work and pushing investment offshore. Both can be true, and neither changes the near-term outcome much: the bills most likely to pass are narrow.

What changes immediately is scrutiny. Committee chairs control hearings, subpoenas and the public record. When a model fails in a visible way, or an agent does something it shouldn’t, the question “what did you know and when” now has a forum. That reaches beyond the AI labs. A bank, insurer or retailer whose AI system harms customers can be asked for its records, its testing and its decisions.

The states still set the rules

The most consequential AI rules for most firms are already written, and they are state rules. California’s and Colorado’s rules on automated decision-making take effect on 1 January 2027. A Democratic Congress is unlikely to preempt them, and Democratic proposals generally treat federal rules as a floor. Any firm that paused state compliance work in the hope of a single federal standard should restart it.

Data centers become a political cost

The election made the cost of computing a local political question. Whether or not a national moratorium passes, and I don’t expect one to, permits, power contracts and tax treatment for new data centers just got harder in some of the biggest markets. That flows through to the price and availability of AI capacity, and to the location of the data centers your providers use. For regulated firms, a provider’s decision to build abroad is a data residency question as well as a cost question.

Evidence is the common thread

Hearings, state rules on automated decisions and incident reporting all ask for the same thing: show what the system did, on what data, under whose authority, and what you did when it went wrong. Firms that can produce that record will find this Congress manageable. Firms that can’t will find every hearing and every state inquiry expensive.

Why it matters

For the industry. Plan for two years of uncertainty at the federal level and steady pressure underneath it. The safe assumption is no federal preemption, more state obligations, more public questions about incidents and rising costs for capacity in some regions. None of that is a reason to stop putting AI into production. It is a reason to build the controls and records now, while they are a choice rather than a demand.

How it connects to our work. AI Governance Evidence sets out the request records and release evidence a reviewer, regulator or committee can inspect. The Data and AI Operating Model covers who decides, who owns each system and how incidents are escalated, which is what a document request will ask about first. Four steps I’d take this quarter:

  • Inventory and tier your AI uses. Know which systems make or shape decisions about customers, and which can act without a person.
  • Keep an incident log. Record AI failures, near misses and fixes as they happen, with the owner and the outcome.
  • Run state compliance to its dates. Treat the January 2027 automated decision rules as fixed, whatever happens in Washington.
  • Ask your providers where they will build. Add location, capacity and price change terms to your AI contracts.

Sources

RegulationAI governancePolicyData centers

Linked sources are the record. Check them for the full reporting.

Read next

Viewpoint · · 6 min read

The OCC will judge bank AI by what it can do, not how clever it is

The OCC’s first guidance for examiners on bank AI ranks systems by the consequence of a mistake, how much authority they hold and how much a person still oversees. A simple agent that can move money can outrank a frontier model that only drafts. Banks will need to show, for each system, what it may reach and do, and who can stop it.

BankingAI governance

Viewpoint · · 5 min read

Where your AI runs is now a business decision, not just your provider’s

Local opposition, power limits and the election have made new data center capacity slower and less certain in some of the largest US markets, and providers are weighing building abroad. For firms putting AI into production, the location, capacity and continuity of the compute behind their AI now belong in the same plan as their data residency and resilience rules.

Data centersAI infrastructure

Viewpoint · · 5 min read

Is the AI bubble bursting? The buildout is wobbling. The demand isn’t.

This week’s tech selloff, a chip CEO calling data centers a bubble and the election backlash against new sites all point to strain in the AI buildout. But the demand data says enterprise use is still growing fast. What is likely to correct is the financing of capacity, not the use of AI, and firms should plan for that difference.

AI economicsData centers

All viewpoints Discuss this with Richard