Trump’s New AI Executive Order Has No Teeth and No Requirements

Here are eight key points summarizing the article:

Executive Order and AI Submission: Under an executive order issued by President Donald Trump, AI companies like OpenAI and Anthropic are requested to submit their advanced AI models to the government for cybersecurity, confidentiality, “insider risk,” and intellectual property protection vetting.
Voluntary Nature: Despite the call for submission, AI companies are not obligated to comply. The order explicitly states that nothing in the section shall create a mandatory licensing or permitting requirement for AI model development or distribution.
Focus on Frontier AI Models: There is particular emphasis on “frontier models”—cutting-edge AI models that pose significant security risks, like the one developed by Anthropic called Mythos, which was withheld due to security concerns.
Competitive Landscape: American AI companies are urging to train and release their models quickly, especially in contrast with the emerging Chinese AI competition.
Concerns Over Regulation: Rapid AI model releases have raised serious concerns that these models might not undergo proper testing or regulation, possibly leading to negative human impacts or misuse as hacking tools.
Government Cybersecurity Initiatives: The executive order also mandates that defense-related agencies such as the Pentagon and Treasury beef up their cybersecurity defenses and have government agencies create a framework within 60 days to evaluate AI models.
Changes from Initial Draft: Expected two weeks before the June 2 order, the initial draft called for a 90-day review period but was shortened to 30 days following pushback from AI companies.
Criticism of Oversight: Despite the executive order, there is a critical lack of enforcement or mandatory requirements, which has led to a muted reaction. Anthony Aguirre of Future of Life Institute argues for a mandatory government pre-deployment review process to block the release of risky AI systems. Similarly, John Thickstun of Cornell University suggests that the order mainly creates an appearance of oversight while maintaining a hands-off approach to AI governance.