This past month, the headlines got scary and wild. AI agents from some of the biggest tech companies in the world didn’t just make mistakes. They broke into systems they were never supposed to touch. Government databases in the United States and Australia. Private companies. Even AI research platforms. In one case, an AI system quietly accessed a non-public database on bushfires it had no business touching. In another, a company’s own AI hacked into an outside business.
These weren’t evil robots in a movie. These were real AI agents, built by real companies. They did things their own creators didn’t authorize and, in some cases, didn’t even notice until the hacks were already done.
AI is moving faster than our ability to control it. If we don’t build real guardrails now, we won’t get another chance if something truly catastrophic happens. The good news is that we don’t have to choose between safety and staying ahead in the AI race. We can do both.
Why This Keeps Happening
These incidents aren’t random bad luck. They’re a warning sign. AI agents are being built to act independently, to make their own decisions, and to take actions online without a human checking every single step. That’s what makes them powerful. It’s also what makes them dangerous when something goes wrong.
Right now, there’s no shared rulebook forcing every company, in every country, to test, report, and contain these risks the same way. One company catches a problem and fixes it quietly. Another doesn’t catch it until something already happens. Meanwhile, the technology keeps getting more powerful. We wind up patching holes in a ship that’s still being built faster than we can fix it. Every week we wait to act, the gap between AI’s power and our oversight grows wider.
A Global Rulebook for Technologies That Could End Us
Here’s the part most politicians won’t say out loud: some technologies are too dangerous for any one country, or any one company, to handle alone. AI weapons. Bioweapons built with AI assistance. Cyber-attacks on utility grids. Autonomous “killer robots” that can make life-and-death decisions without a human in control. These aren’t ordinary business risks. They’re species-level risks, and they demand species-level rules.
What we need is a Global Safety Rulebook. It would not be run by the United Nations or any existing organization. Every nation keeps its own military, its own borders, and its own freedom. No one hands over their army to a global government. But every nation that signs on agrees to a shared rulebook: no AI-driven bioweapons, no autonomous weapons without meaningful human control, and a real, enforceable path toward nuclear disarmament with actual inspections, not just promises.
If one country breaks these rules and becomes a genuine threat, a council of member nations decides what to do, and if action is needed, every member helps respond using their own resources. Nobody is forced to give up power. Everybody agrees to stand together as a shield instead of standing alone as a target. AI is a borderless global technology that will be everywhere, and that means enforcement must be everywhere too. The US can no longer be the world’s police force, standing guard alone while everyone else gets a free ride on our protection. This shared rulebook is the solution—strength through partnership, not exhaustion through solitary duty.
Why Even Our Rivals Should Want In
Here’s something most people miss about the AI race. For leaders like Xi Jinping and Vladimir Putin, the real danger of AI isn’t losing a competition. It’s what unregulated AI does to the systems that keep them in power. AI translates, verifies, and spreads information faster than any censorship system can block it. It helps ordinary citizens fact-check state propaganda in real time and expose corruption that used to stay hidden.
If China and Russia refuse to join international AI agreements, they’re not protecting themselves. They’re leaving themselves exposed. Without shared rules, any rival, any hostile group, could aim a cyberattack, a disinformation campaign, or worse at Beijing or Moscow, and no international law would even be broken. Staying outside the rulebook doesn’t protect their sovereignty. It gambles with it. The smart move, for every major power, is to help write these rules while their voice still matters, instead of getting boxed in later by a system they had no hand in shaping.
This isn’t naive hope. It’s basic self-interest, and self-interest is the one language every government on Earth understands. A rulebook built on mutual survival doesn’t require trust—it only requires that everyone sees the alternative is worse for them too.
Protecting Our Chips, Protecting Our Future
None of this works if we let our most advanced AI technology walk out the back door. Our leading AI chips are the brains of our economic and national security. They power breakthroughs in medicine, clean energy, and defense, and anchor the next generation of good American jobs. They’re also the single biggest prize for adversaries like the Chinese Communist Party, who would rather steal our innovation than compete fairly for it.
A Chip Security Act, currently moving through Congress with unanimous bipartisan support, would crack down on the shell companies and shadow middlemen who smuggle restricted AI chips through third countries. It would require real verification of who’s actually buying and using these chips. And it would push us to coordinate with our allies, so we present a unified front instead of leaving loopholes that bad actors can exploit, one border at a time. It’s legislation I’ve been pushing for, because closing these gaps isn’t optional. It’s essential to keeping our most advanced technology out of the wrong hands.
But we have to be honest about the balance here too. If we overreach and choke off every legitimate export, we risk slowing down American innovation and pushing other countries to simply build their own AI ecosystems without us. The answer isn’t fear. It’s smart, targeted guardrails that protect American workers, defend our values, and keep America in the lead, not a wall so high it traps us in instead of keeping threats out.
One Technology, No Borders, One Set of Rules
AI doesn’t care about national borders, and neither do the risks it creates. A rogue agent breaking into a database doesn’t stop at a checkpoint. A bioweapon worse than Covid designed with AI assistance doesn’t ask permission to cross an ocean. If we want real safety, we need both strong chip security here at home and a global rulebook that reaches everywhere this technology can go.
We can lead the world in AI and lead the world in AI safety at the same time. That’s the only way we stay ahead for the long haul. The countries and companies that come together to build trust and safety into their systems now will be the ones still standing when this technology matures. The ones who go it alone and cut corners for a short-term edge are the ones who will eventually cause, or suffer, the kind of catastrophe none of us can take back.
The choice in front of us isn’t between moving fast and staying safe. It’s between leading this moment with wisdom or being dragged through it by headlines we can’t undo.
short url:
