President Donald Trump signed a new executive order on artificial intelligence Tuesday, establishing a voluntary system for tech companies to share frontier models with the government for security review. The policy, which drops a mandatory licensing requirement, represents a calibrated step toward federal AI oversight after the White House scrapped a stricter predecessor last month. Companies will be asked to submit models 30 days before planned release.
The order creates a dedicated AI cybersecurity clearinghouse to coordinate security checks between government agencies and the private sector. MIT Technology Review first reported details of the policy shift. The New York Times and Reuters also confirmed the signing Tuesday.
The 30-day review window is a significant reduction from the 90-day period requested in the earlier executive order that Trump shelved in May. That earlier version never took effect. The new framework asks rather than compels. "It's a slimmed-down version," the Wall Street Journal noted in its analysis.
NPR reported that companies face no penalties for refusing to participate. The government will not require permits before software deployment. Here is what they are not telling you.
The voluntary nature of the review system masks a strategic calculation. By offering cooperation instead of demanding it, the White House avoids a direct confrontation with Silicon Valley while still establishing a federal foothold in AI safety testing. The clearinghouse becomes the institutional beachhead.
Quay Barnett understands this kind of calibrated pressure. The former Army Special Operations Command officer now leads Anduril's effort to build an augmented-reality headset for soldiers with Meta. His team envisions drone strikes ordered through eye-tracking and voice commands.
The human becomes the weapons system. That vision of military AI integration runs parallel to the civilian oversight questions Trump's order raises. Anduril shared new details about the prototype this week, MIT Technology Review reported.
Soldiers and drones would see together, share information, and decide as one. The defense applications underscore what is at stake in AI governance. Systems that can order lethal action through a glance demand scrutiny.
The executive order's cybersecurity clearinghouse will handle some of that coordination. But the voluntary framework leaves gaps. Meta faces its own internal pressure over AI.
The company scaled back plans to track workers' clicks and keystrokes for AI training after fierce backlash, Reuters reported. Employees can now pause tracking for 30 minutes. Some are fully exempt.
The Information first reported the changes. Microsoft wants users addicted to its new AI assistant. Internal documents for the "Scout" tool revealed the objective, 404 Media reported.
The assistant launched Tuesday, according to TechCrunch. These corporate moves form the backdrop to Trump's order. The White House is attempting to govern a technology that companies are simultaneously racing to deploy and struggling to control.
Mathematicians fear the consequences. A new declaration raises concerns about AI's trustworthiness in the field, Ars Technica reported. It arrived a week after OpenAI claimed to have solved a famous math problem.
The Wall Street Journal covered that announcement. Scientists have found ways to supercharge computer worms with AI. The New York Times reported the worm could target any known flaw in the world's computers.
The threat landscape expands as oversight frameworks remain voluntary. Google must now let UK publishers opt out of AI search features. The BBC reported online publishers can choose not to appear in AI Overviews.
Google is testing features for sites to exit AI search entirely, Reuters confirmed. The European Parliament is going further. It is ditching Google for Quant, a French search engine, as the default on in-house computers.
Politico reported the switch. The Financial Times framed it as part of a broader push to wean the EU off American technology. SpaceX plans to raise $75 billion in an initial public offering at $135 per share.
The company intends to sell 555.6 million shares, Reuters reported. The fixed price breaks from the traditional IPO process, Bloomberg noted. Morningstar says the valuation should be nearly 50% lower, according to Business Insider.
Dan Coatsworth, head of markets at AJ Bell, told CNBC the price looks overloaded with expectations. The math does not add up. America's data center build-out is falling behind schedule.
Sixty percent of facilities planned for completion in 2027 are not yet under construction, the Wall Street Journal reported. Nobody wants a data center in their backyard, MIT Technology Review noted. Electric vehicles are getting cheaper worldwide.
Except in the United States. Rest of World reported the US lacks supportive policies and access to affordable Chinese EVs. In Marseille, activists are fighting back against surveillance.
Eda Nano, a 39-year-old developer, points out what looks like a streetlamp on the Rue des Abeilles. It is a video camera with a 360-degree view. She wants residents to know they are being watched.
MIT Technology Review profiled the growing resistance in the rebellious port city. Why It Matters:
The executive order establishes a federal infrastructure for AI oversight without triggering the political and legal battles that mandatory licensing would provoke. Companies gain a seat at the table. The government gains access to frontier models before they reach the public.
The clearinghouse is only as strong as the participation it attracts. The order also signals to allies and adversaries that Washington is engaging with AI safety, even as Europe moves toward harder mandates. The EU's switch away from Google and the UK's publisher opt-out rules show regulatory divergence accelerating.
Trump's order positions the US somewhere between laissez-faire and the European approach. Key takeaways: - Tech companies are asked, not required, to share frontier AI models with the government 30 days before release. - The order drops mandatory licensing and reduces the review window from 90 days to 30. - The policy marks a strategic shift from the administration's previous hands-off approach to AI governance. What comes next is a test of voluntary compliance.
The clearinghouse needs major AI developers to submit models for review. Without participation, the framework becomes symbolic. Congress may face pressure to legislate if voluntary cooperation fails.
The EU's regulatory trajectory provides a counterpoint. Watch whether any major AI company publicly refuses to participate in the review system. That refusal would define the limits of voluntary governance.
Key Takeaways
— - Tech companies are asked, not required, to share frontier AI models with the government 30 days before release.
— - A new cybersecurity clearinghouse will coordinate security checks between agencies and the private sector.
— - The order drops mandatory licensing and reduces the review window from 90 days to 30.
— - The policy marks a strategic shift from the administration's previous hands-off approach to AI governance.
Source: MIT Technology Review
