The U.S. government just took a big step toward overseeing artificial intelligence. President Donald Trump signed a new executive order asking AI companies to share their most advanced models with the government before releasing them publicly.

This process is voluntary. Companies can choose to submit their AI models up to 30 days before launch. The goal is to let federal agencies test these models for cybersecurity risks. It’s a shift from Trump’s earlier hands-off approach to AI regulation.

Why the change? Security fears grew after Anthropic, a major AI developer, released its Mythos model. Mythos can detect thousands of software vulnerabilities. Some of those flaws affect every major operating system and web browser. That raised alarms about the potential risks of powerful AI tools.

The executive order directs several key agencies—including Treasury, Defense, Commerce, and Homeland Security—to coordinate these reviews. It also sets up a cybersecurity clearinghouse. This new hub will scan AI software for weaknesses and share findings with the private sector.

Importantly, this order does not require companies to get licenses or government approval before releasing AI models. It’s meant to encourage cooperation without stifling innovation. Tech giants like Google, OpenAI, and Anthropic have praised the move and agreed to cooperate.

Balancing Innovation and Security

Trump originally planned a stronger executive order that would have required companies to submit models 90 days before release. But he postponed it, worried it might hurt U.S. competitiveness with China. The current version is lighter but still signals a more active role for the government.

This approach tries to strike a balance. It promotes innovation but acknowledges AI’s growing risks. The government wants to protect critical infrastructure like banks, emergency services, and hospitals from AI-powered cyberattacks.

The order also reflects a broader shift in the tech world. AI is advancing fast, and its risks are becoming clearer. For years, companies mostly policed themselves. Now, governments want a seat at the table to keep AI safe for everyone.

Smart Glasses and the Military Future

On a related front, defense tech companies are exploring how AI can change warfare. Anduril, a defense contractor, is teaming up with Meta to develop augmented reality smart glasses for soldiers. These headsets will let soldiers control drones with eye movements and voice commands.

The goal is to make soldiers and drones work as one system. The idea is “the human as a weapons system,” where information flows instantly between humans and machines. This cyborg-inspired vision could change battlefield decision-making and improve soldier safety.

Anduril’s lead on the project, a former Army Special Operations officer, believes this technology will make soldiers smarter and more connected. Imagine ordering a drone strike just by looking at a target. The technology is still in development, but it points to the future of military AI.

These smart glasses tie into the broader AI security concerns. As AI tools get smarter, they also become more powerful weapons. That makes government oversight and cybersecurity even more urgent.

Overall, the new AI order and military tech projects show how AI is no longer just a tech industry problem. It’s a national security issue, a tool for warfare, and a field demanding careful management. The U.S. government is stepping up, but the balance between innovation and control will stay tricky.

As AI models grow more advanced, expect more policies, debates, and technologies aimed at keeping this powerful tool safe and useful. The next few years will be critical for shaping how AI impacts society and security.