top of page

Has anything changed for AI after Trump’s first week?

Writer's picture: Caro RobsonCaro Robson

28 January 2025


In an historic first, President Trump signed over 20 executive orders and rescinded 78 of his predecessors’ on the first day of his presidency. The new administration also led initiatives on AI infrastructure, energy and free speech in its first week.


But have any of these actions made a real difference to the world of AI?


Today’s post looks at some of the major developments of Trump’s first week and asks whether they are genuinely ground-breaking moves, or just continuations of policies from the Biden administration. I’ll focus on:


  • Trump’s AI Executive Order

  • Rescinding Executive Order 14110

  • The Stargate announcement

  • Free Speech Executive Order


 

Summary


Many of the orders and actions simply continue policies announced during Biden’s term. However, there are three major changes:


  • the shift from environmental concerns

  • the potential removal of federal-level AI guardrails

  • and the chilling effect of an order aimed at promoting free speech


All three changes may have major implications for federal spending and corporate boardrooms across the US. All represent significant divergence with the EU in its digital policy agenda.


Where this will leave EU enforcement of its digital rights package (including the DSA, DMA and AI Act) against tech companies less inclined to comply with such a divergent legal regime is unclear. The possible impact on EU trade relations with the US is also uncertain, as both see AI as key to sustained economic growth.


However, this week also saw China’s DeepSeek produce an AI assistant that outperforms its US rivals whilst relying on less advanced microchips and using less energy. The Stargate project announcement may end up being the last of the mega-compute AI projects, as global stock markets sell-off US tech shares.


Whilst Biden’s microchip embargo may have forced the development of this more efficient model in China, the call to produce more efficient algorithms in his last AI Infrastructure Order may become US policy by necessity, not design.


 

 

Executive Order on Removing Barriers to American AI Innovation


On 23 January, Trump passed his Executive Order on Removing Barriers to American AI Innovation (Trump AI Order), confirming the revocation of Biden’s Order on Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence (Order 14110) and introducing the new administration’s AI priorities:


Sec. 2. Policy. It is the policy of the United States to sustain and enhance America's global AI dominance in order to promote human flourishing, economic competitiveness, and national security.

Executive Order on Removing Barriers to American AI Innovation


The emphasis on “human flourishing” (rather than human rights) is very similar to the language of Silicon Valley. The economic and national security aspects are unchanged from the previous administration, and Trump was predicted to retain them.


There are differences with Order 14110, however.


Order 14110's purpose was to “advance and govern the development and use of AI in accordance with eight guiding principles and priorities” including safety and security, promoting responsible innovation, competition and collaboration, supporting American workers, advancing equity and civil rights, protecting the interests of American users and managing risks from federal government’s use of AI (Section 2). These priorities are not found in the Trump AI Order.


Perhaps a more concerning shift is in the overarching purpose of the Trump AI Order (Section 1): “[The US] must develop AI systems that are free from ideological bias or engineered social agendas.


For anyone comparing the language of the Trump AI Order to the EU’s AI Act, the removal of bias may seem to align with EU regulation. However, in combination with Orders on the reversal of Diversity Equality and Inclusion (DEI) policies and the Order on Free Speech discussed below, this section indicates a shift away from the anti-discrimination bias checks required under the EU’s AI Act.


But what will the Trump AI Order actually do?


Section 4 sets out the core actions under the Trump AI Order:

  • Develop an Artificial Intelligence Action Plan to give effect to the Policy in Section 2 within 180 days

  • The plan will be prepared by the Assistant to the President for Science and Technology (APST), Special Advisor for AI and Crypto and Assistant to the President for National Security Affairs (APNSA) and the heads of various executive departments and agencies

  • The APST is likely to be confirmed as Michael Kratsios, who occupied a similar position in Trump’s first term. Kratsios has been managing direction of an AI infrastructure company, Scale AI, in the intervening period  


Harnessing federal purchasing power…but not safety


The Trump AI Order also seeks to review two of the Biden administration’s key federal AI safety measures.  


Section 4 states: “Within 60 days of this order, the OMB Director, in coordination with the APST, shall revise OMB Memoranda M-24-10 and M-24-18 as necessary to make them consistent with the policy set forth in section 2 of this order.


The OMB (Office of Management and Budget) Memoranda M-24-10 (March 2024) and M-24-18 (September 2024) were the key Biden instruments to ensure that federal government procurement and use of AI was subject to risk management, privacy controls and safety guardrails.


How they will be amended by the new administration is unclear; President Trump has both criticised the Biden administration as imposing too much regulation, but also prioritises national security for AI.


The next 180 days will see the new administration’s federal regulatory policy on AI taking shape, and how the unpicking of all actions under Biden’s Order 14110 will take effect.


 

Rescinding Biden’s Executive Order 14110


On his first day in office, President Trump rescinded 78 of the Biden Administration’s Executive Orders, including Order 14110. The White House website now shows a 404 error message in its place.


Notably, however, Trump’s AI Order did not rescind the AI Infrastructure Order passed on 14 January.


Whether this will also be revoked in the coming weeks remains to be seen, though its provisions on increased energy production may sit well with Trump’s declaration of a national energy emergency.


"I'd like to see federal lands opened up for data centers. I think they're going to be very important"

 

One aspect of the Biden AI Infrastructure Order that may remain is the call to tech companies to produce more efficient algorithms. Section 9 created a “grand challenge” for AI companies, including “developing best practices and standards for software and algorithmic efficiency” (Section 9(k)(v)).


Whilst this may seem at odds with the Trump administration’s energy policy, the recent market reaction to China’s DeepSeek AI assistant may result in this aspect of Biden’s order being retained, if only in practice and not law. DeepSeek was developed using less-advanced microchips as a result of the US embargoes and requires less energy to operate than its US competitors. Global markets have reacted with huge sell-offs of US tech stock. Market reality may end up dictating policy in this area.


How the rescinding of 14110 will happen


The Rescissions Order states that: “A list of additional orders, proclamations and memoranda to be rescinded will be submitted to the President within 45 days.”


Further information on the actions to be taken to rescind Order 14110 can be found in Trump’s AI Order at Section 5 on “Implementation of Order Revocation”:


  • The APST, Special Advisor for AI and Crypto, and the APNSA (Assistant to the President for National Security Affairs) will review “all policies, directives, regulations, orders, and other actions taken pursuant to the revoked Executive Order 14110”

  • The review will “identify any actions taken pursuant to Executive Order 14110 that are or may be inconsistent with, or present obstacles to, the policy set forth in section 2 of this order”  

  • “For any such agency actions identified, the heads of agencies shall, as appropriate and consistent with applicable law, suspend, revise, or rescind such actions, or propose suspending, revising, or rescinding such actions”


Whether the review extends to the numerous NIST guidance documents produced pursuant to Order 14110 - or even results in their revocation – seems very unlikely but may be a possibility. NIST guidance is at the core of US AI standards (its webpages on activities under Order 14110 now simply say that it was rescinded).


The US AI Safety Institute (AISI) was created by NIST the day after Order 14110 was signed. Its future may now be uncertain, although retention of the AI Infrastructure Order (which gave the AISI some oversight functions) may be a sign that it will stay.


Some of the numerous actions taken under Order 14110 include:


  • Using the Defense Production Act to compel developers of the most powerful AI systems to report ‘vital information’ to the Department of Commerce 

  • Proposing a draft rule to compel US cloud companies providing cloud services for training AI to non-US organisations to report this to the Department of Commerce

  • Receiving completed AI risk assessments for critical infrastructure from nine US government agencies as the basis for future federal action on AI safety


All these actions would seem to fit within the remit of protecting national security: a priority for the Trump AI Order. However, they may be seen as too aggressive regulation for the new administration.


How the Trump Administration plans to unpick all the initiatives, rules and actions taken under Order 14110 is unclear. It will be a huge task, potentially dismantling federal US AI safety regulation.


The result may be that AI regulation becomes more of a state, rather than a federal, issue. Four states already have comprehensive AI laws, with over 45 others having legislation that touches upon AI. In a great article on 24 January, Forbes argued that the new AI orders will increase, rather than decrease, the pressure on boards to implement AI safety regimes, as it will result in greater uncertainty. Again, the next 180 days may be key in understanding the new administration's approach.

 

 

Stargate Project Announcement


On 22 January, Trump announced the Stargate Project, a $500 billion joint venture between OpenAI, SoftBank and Oracle, with investment firms from Japan and the UAE, to build “the largest AI infrastructure project in history so far.”


The project was first made public in April 2024, with The Times reporting that the original concept was for Microsoft and OpenAI to fund a $100 billion data centre, using around 5GW of energy (more than Greater London uses in a year). The project announced last week is already being built in Texas, with costs having grown to $500 billion and Oracle being the key infrastructure partner in place of Microsoft.


Stargate may be a project from the Biden era, but its potential environmental impact could be pure Trump if its energy consumption is not met from renewable or nuclear energy sources.


However, with China’s DeepSeek shaking global markets this week, Stargate may be outdated even before it is built. DeepSeek R1 is an AI assistant that uses less energy and less-advanced microchips than its US competitors, relying instead on algorithmic efficiency as a result of the US trade embargoes that have reduced access to the most advanced chips in China.


DeepSeek is open source, but reports indicate that it tracks keyboard strokes (like TikTok), suggesting that it could be seen as a national security threat. However, with Trump intervening to pause the ban on TikTok in the US, it may be that his administration feels unable to prevent DeepSeek from entering the US market. If DeepSeek can genuinely out-perform more power-hungry, superprocessor-reliant competitors, the market may end up determining AI energy policy rather than the White House.


A further potential ramification is the relationship between President Trump and his close advisor Elon Musk. Musk and OpenAI CEO Sam Altman founded OpenAI together, but have since become bitter rivals, with Musk suing Altman over the company’s shift to a for-profit structure. Musk was quick to condemn the Stargate announcement; within minutes he had posted on X that funding was not secure. How the President responded to this interjection is not known, but it could signal a rift between him and his new head of the Department of Government Efficiency (DOGE).

 

 

Executive Order on Free Speech


On his first day in office, Trump passed his Executive Order on Restoring Freedom of Speech and Ending Federal Censorship (Free Speech Order). The Order makes reference to the US Constitution, appearing to reaffirm First Amendment rights to free speech. However, the actions ordered are more concerning.


Over the last 4 years, the previous administration trampled free speech rights by censoring Americans’ speech on online platforms, often by exerting substantial coercive pressure on third parties, such as social media companies, to moderate, deplatform, or otherwise suppress speech that the Federal Government did not approve.  

Free Speech Order, Section 1


In addition to pushing against misinformation and content regulation, the Free Speech Order also requires an investigation into past government action. Section 3(b) orders the Attorney General to “investigate the activities of Federal Government over the last 4 years that are inconsistent with the purposes and policies of this order” and prepare a report with “recommendations for appropriate remedial actions to be taken.”


Although the “remedial actions” are aimed at the Federal Government, there are clear implications for social media platforms. Meta reduced its fact-checking and content moderation teams drastically prior to the inauguration, with reports from The Rest is Politics US that Mark Zuckerberg is actually afraid of being arrested for prior content control on his platform.


In an irony George Orwell may have appreciated, the Free Speech Order may actually have a chilling effect on free speech if platforms feel compelled to allow misinformation and remove their control over content moderation.  


However, this federal restriction on content control was foreshadowed by the July decision of the Supreme Court in Moody v Netchoice, which remitted cases back to federal courts in Texas and Florida that challenged state laws preventing platforms’ content moderation. The Court held that neither state court had properly considered the First Amendment (free speech) ramifications of preventing states from banning content moderation.


If the US imposes an effective ban on content moderation at the federal level, this could bring it into further conflict with the EU, whose Digital Services Act seeks to compel social media platforms to carry out rigorous content control. Platforms may be required to operate entirely different versions in the US and EU as a result.

 

 

Conclusion: Has Trump’s first week made a real difference to AI?


The answer to this question is: sort of.


On the one hand, the rescinding of Order 14110 has the potential to remove federal AI guardrails, making AI safety a state issue. The ensuing regulatory complications for boardrooms could require substantial investment in compliance teams, assuming state regulators will enforce local laws. The Free Speech Order also has the potential to make content moderation effectively unlawful in the US, in sharp contrast to the EU which now mandates it.


On the other hand, the Stargate project was underway before 20 January and may be superseded by more efficient AI systems from China. Microsoft’s Satya Nadella told Davos this week that these developments must be “taken seriously.” It may be, therefore, that market forces shape the future of the US AI industry more than government policy.


One immediate impact, however, is the reaction of foreign leaders. In the UK, the Chancellor has announced a major programme of deregulation, pushing for a “UK Silicon Valley” to be built around Cambridge, approving a £4 billion data centre, and removing the Competition and Markets Authority chairman for a former Amazon executive in a move to reduce regulatory oversight. The Chancellor told reporters that her policy was still essentially “Bidenomics” but that the “framing” will change now that President Trump is in power.


It is tempting to see this as the impact of Trump’s first week in power: essentially the same, but with different “framing.” However, the Free Speech Order and possible removal of federal guardrails will have significant impacts on the digital and AI space. Even so, as DeepSeek’s rise demonstrates, it may ultimately be technology and the market – not the White House – that determine US AI policy.

bottom of page