Across the technology industry, engineers and researchers who build advanced AI systems are arriving at a shared conclusion: the risks these systems pose have outgrown the governance structures meant to contain them. Rather than calling for a halt to development, they are urging the United States to lead a binding multilateral effort that would establish common standards before fragmented national rules create dangerous gaps. It is a rare moment when those closest to a powerful technology are asking, with some urgency, to be governed more carefully.
Tech Workers Push for US-Led Global AI Risk Management Framework
A framework built today might shape how the field develops
Mark
Why are tech workers specifically pushing for US leadership on this? Why not a neutral international body?
Mimi
Because the US has the credibility and the leverage. If China or Russia led the effort, Western companies wouldn't trust it. If the UN tried, it would move too slowly and lack enforcement teeth. The US can actually get other countries to the table.
Mark
But doesn't that just mean the US is trying to lock in its own advantage?
Mimi
Maybe. But these workers aren't saying the US should write the rules unilaterally. They're saying the US should convene the process. The actual standards would have to be negotiated.
Mark
What happens if countries don't agree?
Mimi
That's the real problem. You end up with some countries playing by strict rules and others not. Companies migrate to the permissive jurisdictions. Safety gets undermined by competition.
Mark
So this is about preventing a race to the bottom?
Mimi
Exactly. Right now, there's no floor. A global framework would establish one—a baseline that everyone has to meet.
Mark
And these workers think that's actually achievable?
Mimi
They think it has to be. The alternative is letting AI development happen in a regulatory vacuum, and they don't believe that ends well.
O Pulso
- Advanced AI is being woven into critical infrastructure, medical systems, and military applications faster than any regulatory body has been able to respond.
- The current patchwork of national rules creates a race to the bottom, rewarding companies that relocate to lighter-touch jurisdictions and leaving safety researchers without shared ground to stand on.
- Tech workers are calling on the United States — citing its technical credibility, economic weight, and diplomatic reach — to convene other nations around a binding, enforceable global framework.
- The window for action is narrowing: a framework shaped now, while the technology is still forming, carries far more leverage than one negotiated after trillions of dollars and entrenched interests have locked the field in place.
- The hardest obstacle is translation — turning worker consensus into actual treaties requires governments that move slowly and nations with competing AI ambitions to find common ground they have not yet sought.
Across the technology industry, engineers and researchers who build advanced AI systems are arriving at a shared conclusion: the risks these systems pose have outgrown the governance structures meant to contain them. Rather than calling for a halt to development, they are urging the United States to lead a binding multilateral effort that would establish common standards before fragmented national rules create dangerous gaps. It is a rare moment when those closest to a powerful technology are asking, with some urgency, to be governed more carefully.
Inside technology companies across the country, a quiet consensus has taken shape among the engineers and researchers who build advanced AI systems: the world needs a single, coordinated approach to managing what these systems can do. They are not asking for a ban. They are asking for something they consider more urgent — a global oversight framework that closes the gaps a fragmented, jurisdiction-by-jurisdiction approach inevitably leaves open.
The core observation driving this push is simple: AI is advancing faster than governance can follow. A company building a powerful system in one country operates under one set of rules; move the work across a border and the regulatory landscape changes entirely. That fragmentation, these workers argue, creates perverse incentives — it encourages regulatory arbitrage, undermines shared safety standards, and leaves spaces where no one is watching.
What they want is US leadership in negotiating a multilateral agreement: binding, enforceable, and built around common rules for how advanced AI is developed, tested, and deployed. The reasoning is pragmatic. America has the technical expertise, economic weight, and diplomatic standing to convene other nations around credible standards. If it moves slowly or steps back, other powers will shape the framework instead — and the result may not reflect the safety priorities these workers believe are most essential.
The challenge is that tech workers can articulate the need but cannot negotiate treaties or enforce compliance. Governments must do that work, and governments move slowly across competing national interests. Whether other nations with their own AI ambitions would accept a US-led framework remains an open question.
Still, those pushing for change argue the moment matters precisely because it will not last. A framework established now, while the technology is still relatively young, can shape how the field develops. One built years from now, after vast capital has been committed and powerful interests have formed, will be far harder to construct and enforce. The argument is not that the risks are new — it is that the opportunity to address them systematically may be closing.
Inside technology companies across the country, a quiet consensus has formed among engineers and researchers: the world needs a single, coordinated approach to managing the risks that advanced artificial intelligence systems pose. These workers are not calling for a ban on AI development. They are calling for something more specific and, in their view, more urgent—a framework for global oversight that would prevent a chaotic patchwork of national rules from fragmenting the field and, worse, from allowing the most dangerous capabilities to slip through the cracks.
The argument these tech employees are making rests on a straightforward observation: artificial intelligence is advancing faster than governance structures can accommodate. A company building a powerful AI system in one country faces one set of rules. Move that same work across a border, and the regulatory landscape shifts entirely. This fragmentation, they contend, creates perverse incentives. It encourages companies to shop for jurisdictions with the lightest touch. It makes it harder for safety researchers to establish shared standards. It leaves gaps where nobody is watching.
What these workers want is for the United States to take the lead in building something different—a multilateral agreement, binding and enforceable, that would establish common ground rules for how advanced AI systems are developed, tested, and deployed. The reasoning is pragmatic rather than idealistic. The US, they argue, has the technical expertise, the economic weight, and the diplomatic standing to convene other nations around a credible set of standards. If America sits out or moves slowly, other powers will fill the vacuum, and the resulting framework may not reflect the safety priorities these workers believe matter most.
This is not a fringe position. It reflects a growing recognition within the industry itself that the current trajectory is unsustainable. Advanced AI systems are becoming more capable and more consequential. They are being integrated into critical infrastructure, medical systems, financial networks, and military applications. The risks are no longer theoretical. They are operational. And the people building these systems—the ones who understand their capabilities and limitations most intimately—are saying that voluntary measures and light-touch regulation are not enough.
The challenge, of course, is translating this sentiment into actual policy. Tech workers can call for international governance frameworks, but they cannot negotiate treaties or enforce compliance. That work falls to governments. And governments move slowly, especially when they are trying to coordinate across borders and competing interests. There is also the question of whether a US-led framework would actually command global participation. Other nations have their own AI ambitions and their own views about what safety looks like. Getting them to agree on binding rules is not a given.
But the workers pushing for this change are betting that the window for action is closing. As AI capabilities advance, the stakes of getting governance wrong grow higher. A framework built today, while the technology is still in relative infancy, might actually shape how the field develops. A framework built five years from now, after trillions of dollars have been invested and powerful interests have formed, will be far harder to establish and enforce. The argument, in essence, is that this is the moment—not because the risks are suddenly new, but because the opportunity to address them systematically may not last much longer.
Citações Notáveis
Tech workers recognize that advanced AI poses significant risks requiring coordinated global oversight rather than fragmented national approaches— Industry consensus
Quer a matéria completa?
Leia o original em Reuters ↗
Technology