What Trump's victory means for AI regulation

Silicon Valley executives and AI developers spent election night holding their breath, and now the entire regulatory playbook in Washington is about to be rewritten.

Following Donald Trump’s victory on November 6 to become the 47th U.S. president, accompanied by a Republican Senate and potentially a compliant House, the federal approach to artificial intelligence is heading for a dramatic U-turn.

Trump spent the campaign trail promising to repeal President Joe Biden’s landmark AI executive order on day one. After more than a year of intense debate over safety benchmarks and reporting rules, that framework looks ready to crumble.

Tearing down the federal safety rails

Biden’s October 2023 executive order never carried the weight of permanent federal statute, relying instead on voluntary participation backed by the 1950s Defense Production Act. Yet its core provisions drew intense backlash from conservative lawmakers and tech advocates.

The most controversial rules required frontier AI labs to report their training methods and hand over red-teaming vulnerability test results directly to the government. At the same time, the Commerce Department’s National Institute of Standards and Technology (NIST) was tasked with authoring standards to weed out algorithmic bias, leading to the creation of the U.S. AI Safety Institute (AISI).

Trump’s allies argued these demands amounted to government overreach that forced startups to expose trade secrets. Representative Nancy Mace warned during congressional hearings that the burdens would scare off innovators and choke off future ChatGPT-scale breakthroughs.

Vice President-elect JD Vance voiced similar concerns, arguing that preemptive regulation would simply entrench incumbents. Meanwhile, critics like Senator Ted Cruz attacked NIST’s safety standards as an attempt to enforce woke ideological control. Trump pledged instead to champion policies supporting AI rooted in free speech, signaling an end to federal scrutiny of model outputs.

With that mandate, the future of the AISI appears grim. Unless Congress moves quickly to codify the institute into permanent law, the body could simply vanish with the stroke of a pen.

Light federal touch meets state fragmentation and tariffs

Walking away from federal oversight does not mean AI teams will operate in a vacuum. If anything, the legal terrain might get messier.

Dean Ball, a research fellow at George Mason University, expects a light-touch regime that leans on existing statutes rather than creating new AI-specific rules. The catch is that states are already racing to fill the vacuum.

Policymakers across the country have introduced nearly 700 AI-related bills this year alone. Tennessee enacted legislation to shield voice artists from AI clones, Colorado rolled out risk-tiered compliance rules, and California Governor Gavin Newsom signed dozens of safety bills compelling companies to disclose training data. If individual states keep rolling out conflicting local rules, founders could face a regulatory nightmare across state borders.

Then there is the trade ledger. Trump campaigned on broad protectionist measures, proposing a 10% tariff on all imported goods and up to 60% tariffs on Chinese imports, alongside tighter chip export bans.

Researchers like Hamid Ekbia of Syracuse University and Matt Mittelsteadt of George Mason warn that these policies could severely disrupt hardware supply chains. Higher compute tariffs could starve labs of the capital needed for core model research, creating an entirely different kind of headwind.

Why it matters

We think Trump’s return to power represents a massive shift in how AI will be built and commercialized, but it is hardly an unmitigated win for founders.

For early-stage startups and open-source builders, the relief is tangible. You will likely not have to spend legal capital navigating onerous federal disclosures or proving compliance with amorphous bias guidelines. But the trade-off is fragmented state laws. If California and New York impose their own divergent requirements, national compliance could actually become more burdensome than a single federal standard.

For incumbents like OpenAI and Anthropic, this upends their Washington playbook. The safety guardrails they spent the past year negotiating were effective tools for shaping the regulatory moat. Without them, the battle reverts to raw compute scale, operational speed, and product execution.

Our view is that discarding hard safety discussions purely on partisan grounds carries real risks. Stripping away reporting on bioweapon proliferation or systemic cybersecurity vulnerabilities might feel like freedom today, but an unchecked federal vacuum paired with skyrocketing hardware costs is a precarious foundation for the next wave of intelligence.