Overview
On January 20, 2025 — his first day back in office — President Donald Trump signed Executive Order 14179: “Removing Barriers to American Leadership in Artificial Intelligence.” The order explicitly revoked President Biden’s October 2023 AI Executive Order, which had required AI companies to share safety testing results with the federal government before deploying large AI models.
The same day, DeepSeek R1 was released in China — a juxtaposition that shaped the entire political framing of US AI policy for 2025.
What Biden’s Order Had Required
Biden’s October 30, 2023 Executive Order on AI — “Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence” — was the most expansive US AI policy action to that point. Key provisions:
- AI companies training models using more than $100M in compute must report safety test results to the government
- Federal agencies must develop AI risk frameworks
- National security review for AI models with potential weapons capabilities
- Guidelines for watermarking AI-generated content
The order was implemented through the National Institute of Standards and Technology (NIST) and relied heavily on voluntary commitment alongside the formal requirements.
What Trump’s Order Did
EO 14179 revoked the Biden order and directed:
- Federal agencies to remove regulatory barriers to US AI leadership
- The development of an “AI Action Plan” prioritizing American dominance (published July 2025)
- Review and removal of existing AI regulations deemed to hamper innovation
- No new federal mandates on AI safety testing before the action plan was finalized
In December 2025, Trump signed a second AI executive order establishing a National AI Policy Framework that:
- Directed the Department of Justice to challenge state-level AI laws that created “onerous” compliance burdens
- Created a federal preemption strategy to prevent a patchwork of 50 state AI regulations
- Directed the FTC to issue an AI policy statement
- Established a DOJ AI Litigation Task Force
The Philosophy Behind the Policy
The Trump administration’s AI approach was articulated clearly: the greatest threat to US AI leadership is not unsafe AI but over-regulated AI. In the context of DeepSeek R1’s release, this translated to: if China can build frontier AI at $6M without US safety requirements, then US safety requirements are a handicap.
This position had vocal support from key technology investors and executives:
- Marc Andreessen called the Biden order “AI safety theater” that gave China an uncontested lane
- Elon Musk (who simultaneously was launching his own AI company, xAI) advocated for deregulation while paradoxically raising concerns about competitor AI safety
- The prevailing framing in Silicon Valley: speed is safety — the US must stay ahead to ensure aligned AI wins
The Counterarguments
The deregulatory turn was contested by:
- Anthropic CEO Dario Amodei: argued that safety and capability are complements, not trade-offs — the most capable AI systems that lack safety properties pose the greatest risks, regardless of who builds them
- Geoffrey Hinton and other AI researchers: the risk of untested deployment is not regulatory burden but civilizational consequence
- EU regulators: the Brussels Effect means US deregulation exports risk to the rest of the world, where American AI systems still operate under EU law
Global Impact
The US deregulatory pivot had international ripple effects:
- EU regulators accelerated implementation of the AI Act, increasingly positioned as the global safety floor
- UK policy shifted toward alignment with Washington, rejecting the Paris summit’s multilateral safety framework
- China continued its parallel track of domestic AI regulation (Generative AI Measures, 2023) while deregulating for international competitiveness
- Three distinct regulatory regimes were now clearly established: US (deregulatory), EU (rights-based mandatory), China (state-directed)