Summary of Key Points
Trump recently signed the executive order "Promoting Advanced Artificial Intelligence Innovation and Security" in a low-profile manner. The main focus of this order is to implement a 30-day voluntary pre-review for cutting-edge AI models with advanced networking capabilities (the original draft proposed 90 days, which was shortened after industry lobbying). The order relies on "industry self-regulation" and does not impose legal obligations, but it does carry a form of "soft constraint." It addresses both the security risks associated with AI and public concerns, while providing clear regulatory expectations for AI companies preparing for IPOs (such as Anthropic and OpenAI), which is seen as positive for the tech industry. The executive order also includes supporting measures such as cybersecurity collaboration and talent recruitment, and it leaves room for potential future regulatory enhancements.
1. Core Changes: 30-Day Review Period + Voluntary Framework to Balance Innovation and Security
The original draft required AI companies to provide the government with access to their models for up to 90 days before releasing them, which worried tech companies about slowing down their product development. After lobbying, the review period has been reduced to 30 days, and it has become a voluntary initiative. Companies can decide for themselves whether their model falls under the category of "cutting-edge models." If so, they can voluntarily grant the government access 30 days in advance and choose trusted partners to participate in the review process.
Companies like OpenAI and Google support this change, with former White House AI director David Sax stating that the 30-day period "will not significantly slow down innovation" as it allows the government to assess security risks without hindering technological progress.
2. Why is It "Soft" Regulation? No Legal Obligation, but Companies Are Likely to Cooperate
The executive order explicitly states that there will be no mandatory approvals. Therefore, companies will not face fines or bans if they do not cooperate. However, lawyer Xiang Chen argues that this is essentially "soft coercion": to avoid potential future issues with the government (such as additional scrutiny), companies are likely to comply voluntarily.
For example, it's like when schools ask students to submit homework voluntarily; although there are no penalties for not doing so, students usually still submit it because they don't want to upset their teachers. Currently, there is no legal obligation, but if Congress passes legislation in the future, it could become mandatory.
3. The Timely Release: A Critical Period for IPOs + Pressure on Security and Public Opinion
The release of the executive order coincides with several important events:
1. IPO Surge: Companies like Anthropic are secretly submitting their IPO applications, OpenAI is preparing for a listing, and SpaceX has started its roadshow (with a potential valuation in the trillions). These companies are particularly concerned about sudden changes in regulatory policies that could affect their valuations.
2. Security Risks Revealed: Anthropic's Mythos model demonstrated advanced networking capabilities, raising concerns among the government and businesses about the potential use of AI in cyberattacks.
3. Deteriorating Public Attitudes: Some Trump supporters are calling for mandatory AI reviews, fearing that AI could lead to job losses or security issues.
By issuing this order, the government is addressing both security concerns and public opinion while providing IPO companies with reassurance about a clear regulatory framework, reducing uncertainty.
4. A Major Boost for IPO Companies: Eliminating Valuation Risks and Opening Up Government Markets
For AI companies preparing for an IPO, the biggest concern is how future regulations might change. This executive order establishes clear rules that are voluntary and do not require mandatory permits, giving the market a clear compliance framework with minimal additional costs.
Analyst Evans notes that this signals the government's willingness to collaborate as a partner rather than an overbearing regulator, reducing risks for companies and their partners. It also opens up potential opportunities for the government to become significant customers of AI models (for example, the Department of Homeland Security may use AI for cyber defense), providing stable revenue for these companies. AI cybersecurity vendors, in particular, could benefit from new "information exchange centers" established under the order.
5. Additional Supporting Measures
In addition to the review process, the executive order includes the following requirements:
1. Cybersecurity Enhancements: The Department of Homeland Security is required to issue guidance within 30 days to strengthen cybersecurity for the government and critical infrastructure, and it will provide AI tools to local governments.
2. Information Exchange Centers: Departments such as the Treasury Department will work with AI companies to establish platforms for sharing vulnerability information and coordinating patch distributions to prevent malicious use of AI.
3. Talent Shortage Resolution: The Office of Personnel Management is tasked with expanding recruitment channels for cybersecurity experts within 60 days, as AI security requires a large number of specialized professionals.
Overall, this executive order represents a balance struck by the Trump administration under the theme of "focusing on development while minimizing regulation." It does not stifle AI innovation but addresses security and public concerns, while also clearing obstacles for companies preparing for IPOs. The future will depend on whether companies actually comply with the requirements and whether Congress enacts more stringent laws. However, for the current AI industry, this is indeed good news.