Donald Trump has spent much of his second term championing a hands-off approach to artificial intelligence regulation. But that stance appears to be shifting slightly. Amid growing concerns about AI’s rapid expansion and safety issues, the president has signed an executive order designed to give the government a chance to review powerful AI models before they are released.
What Does Trump‘s AI Oversight Executive Order Say?
President Donald Trump signed an executive order titled “Promoting Advanced Artificial Intelligence Innovation and Security” on June 2, 2026. The order establishes a voluntary framework for the federal government to evaluate the risks posed by advanced AI systems before they are released, encourages AI companies to share frontier models for security reviews and directs federal agencies to strengthen efforts around AI-related cybersecurity threats. At the same time, it explicitly states that the process is voluntary and does not create any government licensing, permitting or approval requirements for new AI models.
Whether or not to impose guardrails around AI has been a hot topic in the United States for a long time, with 80 percent of U.S. adults prioritizing rules for data security and AI safety over fast-paced innovation, according to a 2025 Gallup poll. But this is the first time Trump seems to be listening to Americans’ anxieties, raising the question of what this could mean for an AI industry that has enjoyed a relatively chummy relationship with the White House thus far.
What Does the Executive Order Say?
Trump’s new executive order, titled “Promoting Advanced Artificial Intelligence Innovation and Security,” is designed to balance two competing goals: keeping the United States at the forefront of AI development while also addressing growing national security and cybersecurity concerns around increasingly powerful models.
The order explicitly states that it does not create any new “mandatory governmental licensing, preclearance, or permitting requirement for the development, publication, release, or distribution of new AI models, including frontier models.” Rather, it establishes a largely voluntary framework for collaboration between top AI developers and the federal government.
Key provisions include:
- Voluntary pre-release reviews for advanced AI models: AI companies can choose to provide the federal government access to powerful new models up to 30 days before they are released publicly. The goal is to identify serious security risks, particularly those related to cyberattacks and national security threats.
- A focus on AI-enabled cybersecurity risks: The order directs national security agencies to develop and maintain a “classified benchmarking process” to evaluate how frontier AI systems could be used to discover software vulnerabilities, conduct cyber operations or otherwise threaten critical infrastructure.
- Creation of a federal cybersecurity coordination effort: The administration plans to establish a centralized “cybersecurity clearinghouse” and improve coordination among agencies responsible for detecting and responding to AI-related threats.
- More government expertise on AI: Federal agencies are instructed to strengthen their AI and cybersecurity capabilities, including hiring and training personnel with relevant technical expertise.
- Preserving U.S. competitiveness: The order repeatedly emphasizes that security measures should not slow innovation or undermine America’s position in the global AI race, particularly against China. Officials involved in drafting the order specifically sought to avoid requirements that could be interpreted as government licensing or approval of AI systems.
A previous draft of the order reportedly called for a voluntary review up to 90 days before release, but industry insiders pushed for something closer to a two-week window. Trump also delayed signing the EO in May, claiming he disliked with “certain aspects” of the proposal and did not want to hinder American AI companies as they compete with China.
“It is the policy of the United States to promote AI innovation and security by working collaboratively with the private sector to modernize government and private sector information systems and harden them against external threats; to protect American ingenuity and intellectual property from exploitation and theft by adversaries; and to cultivate America’s advanced AI-enabled capabilities,” the executive order stated. “As these capabilities evolve, my Administration will continue to work closely with industry to ensure that the best and most secure technology is deployed rapidly to confront any and all threats to our country. We will continue to lead an America First cybersecurity effort that enhances both our national security and our global AI dominance.”
What Has Trump’s Track Record on AI Regulation Been Thus Far?
Even before reclaiming the presidency, Trump distinguished himself as a candidate by promising to deregulate AI. And so far he has made good on his promise. Shortly after returning to office, he repealed former President Biden’s executive order that aimed to set guidelines around AI safety, replacing it with his own order to remove “barriers” to AI development in the U.S.
Trump’s initial attempt to prevent further AI regulations was thwarted when the U.S. Senate removed a key provision banning state laws on AI from his “One, Big, Beautiful Bill” Act. He quickly followed this up by releasing an AI Action Plan that called for scaling back federal regulations to spur the construction of new data centers and ramp up American AI exports. Taking it a step further, Trump signed an executive order at the end of 2025, directly threatening states that planned to maintain their rules on the technology.
These actions set the stage for Trump to propose a national AI framework that would replace any state laws with a single federal rule. The problem is that the framework offers little guidance on how the government should regulate the technology. While it touches on topics like children’s digital privacy, copyright laws and workforce training, much of the language is vague and fails to specify the steps Congress should take to implement the framework’s measures
In short, Trump has largely sacrificed guardrails in the name of spurring on American tech innovation to win the global AI race. That makes his pivot to government oversight (no matter how slight) even more surprising, reflecting just how rapidly the American AI industry is progressing.
Why Has Trump’s Stance on AI Regulation Shifted?
The AI industry entered new territory when Anthropic announced Claude Mythos, its most advanced large language model yet. Possessing unprecedented coding and reasoning abilities, the model exposed tens of thousands of software vulnerabilities, including many that had gone undetected for decades. At this rate, revealing these bugs all at once would make it impossible for software teams to perform patches before they’re exploited. Rather than chance a nightmare scenario, Anthropic chose to hold off on publicly releasing Mythos for now.
Although the Pentagon had already blacklisted Anthropic for refusing to grant unconditional access to its Claude models, the noise around Mythos became too loud to ignore. Trump officials met with CEO Dario Amodei to discuss how the model could support national security efforts, potentially opening the door to reconciliation.
That doesn’t mean that Trump approves of Mythos being released broadly. In fact, the cybersecurity risks posed by the model may be the tipping point that forces Trump to embrace a more hands-on approach with artificial intelligence.
What Could This Mean for the AI Industry?
The initial news that Trump was considering an executive order like this immediately drew positive reactions. A national poll by Morning Consult found that 47 percent of Republican voters “strongly support” Trump’s plan to test AI models before public release, compared to just 5 percent who “strongly oppose” it. More than 70 percent also believe in legally mandatory testing done by an independent party. Meanwhile, another poll by the Institute for Family Studies found that more than 80 percent of Americans supported Trump’s plan.
The backlash against AI has become a rare unifier across the political spectrum, culminating in data center moratoriums, pro-regulation super PACs and even a letter from Congress urging Trump to address the security threats posed by advanced models like Mythos. With the 2026 midterms fast approaching, Trump may feel the pressure of Americans’ discontent with AI and decide to draw harder boundaries around the technology.
For their part, tech companies are aware that their models present new dangers and are already taking steps to mitigate the risks. Anthropic released Claude Mythos only to a set of trusted partners for initial testing, given the security concerns surrounding it. And OpenAI did the same when unveiling its GPT-5.4-Cyber model.
Limited rollouts will likely become the standard among AI leaders, but they should also expect their models to undergo federal testing in the near future. These measures are bound to be viewed as necessary as the AI race heats up, especially in light of U.S. companies accusing their Chinese counterparts of illegally using their models to train their own. Artificial intelligence has officially become a matter of national security, making it essential that Trump adjust his approach to balance tech innovation with the appropriate safeguards.
Frequently Asked Questions
What has Trump’s stance been on AI regulation?
President Trump has promoted a deregulatory vision for artificial intelligence in his second term. After repealing former President Biden’s executive order on AI safety, he proposed an AI Action Plan that called for fewer rules and passed an executive order attempting to ban state laws on the technology. In 2026, he passed an AI framework designed to replace state laws with a single federal rule, although it contained vague language that was light on details.
Which AI companies could be affected by this executive order?
The executive order does not mention any AI companies by name, but it is primarily aimed at companies developing the most advanced AI systems, often referred to as frontier models. That could include firms such as OpenAI, Anthropic, Google, Meta and xAI. Because the framework is voluntary and focuses on highly capable models that could pose cybersecurity or national security risks, smaller AI startups and companies building more specialized tools are less likely to be affected.
Can the federal government stop an AI model from being released under Trump's executive order?
No. Th executive order explicitly states that it does not create a mandatory licensing, permitting or approval process for AI models. Companies can voluntarily submit their advanced systems for government review, but the order does not give federal agencies the authority to block or delay a model’s release. Any broader restrictions on AI development or deployment would likely require new legislation from Congress or action under existing laws.
