As artificial intelligence grows more powerful, the companies building it have begun constructing the walls meant to contain it — a self-regulatory body modeled on financial oversight, funded by the industry itself. This move arrives against a backdrop of urgent warnings from former researchers and a Washington that has chosen, for now, to look away. The tension between those who see existential danger and those who call those dangers a hoax reflects a deeper civilizational question: who decides the rules for technologies that may outgrow any single authority's ability to govern them.
AI Giants Push Self-Regulation as Trump Dismisses Safety Concerns
The dispute is over who writes the guardrails, not whether they're needed
Mark
So these companies are essentially saying they'll police themselves. How does that actually work?
Mimi
They're modeling it after FINRA, the financial regulator. The idea is that industry funding would support independent testing of the most powerful AI systems before release. The companies themselves would contribute the money.
Luke
But here's the thing—we don't know yet if this body will have real teeth or if it's mostly theater. And we should be clear: this is a proposal. It's not operational yet.
Mark
What's driving them to do this now, if Trump is dismissing the whole concern?
Mimi
The warnings from inside the industry itself. Two former researchers just went public saying AI could pose existential risks. That kind of insider alarm tends to move companies, even when politicians won't.
Luke
Though it's worth noting those are two voices. We don't have a survey of what most AI researchers think. The industry is divided on how serious the risks actually are.
Mark
And Congress just isn't moving on this?
Mimi
Not at all. Trump's called it a hoax, and his party controls Congress narrowly. The midterms are coming, so there's no political appetite for regulation right now.
Luke
We should be precise: Congress is divided, but the Republican majority isn't pushing for regulation. That's the actual situation. Some Republicans like Vance have expressed caution, but they're not driving policy.
Mark
What about the smaller AI companies? They seem upset.
Mimi
They're worried that the big labs—OpenAI, Google, Anthropic—are using safety concerns as cover to lock in their market position. Cohere's CEO called it a cartel.
Luke
That's his accusation, but we don't have evidence yet that the regulatory body would actually work that way. It's a real concern worth watching, but it's still speculative at this stage.
Le Pouls
- Former AI researchers are sounding alarms in unusually stark terms — one warning the technology could 'kill us all,' another accusing labs of 'gambling with our lives.'
- The Trump administration has dismissed these warnings as hoaxes, leaving a regulatory vacuum that neither the White House nor a divided Congress shows any urgency to fill.
- OpenAI, Anthropic, and Google are moving to fill that vacuum themselves, building an industry standards body modeled on financial regulation to test the most powerful AI systems before public release.
- Smaller AI companies fear this self-regulatory framework is less a safety net and more a drawbridge — one that locks in the dominance of the labs powerful enough to write the rules.
- With midterm elections approaching and Washington's attention elsewhere, meaningful government oversight remains a distant prospect even as the technology accelerates.
As artificial intelligence grows more powerful, the companies building it have begun constructing the walls meant to contain it — a self-regulatory body modeled on financial oversight, funded by the industry itself. This move arrives against a backdrop of urgent warnings from former researchers and a Washington that has chosen, for now, to look away. The tension between those who see existential danger and those who call those dangers a hoax reflects a deeper civilizational question: who decides the rules for technologies that may outgrow any single authority's ability to govern them.
The world's largest AI companies are moving to police themselves. OpenAI announced Tuesday that it is joining Anthropic and Google in building an industry standards body to oversee the most powerful AI systems before they reach the public — a model borrowed from financial regulation and funded by the sector itself. The idea originated with Google DeepMind's Demis Hassabis, who proposed in July that such a body should have the technical talent and computing resources to conduct rigorous evaluations.
The push comes as warnings from within the industry have grown unusually urgent. Jacob Coxon, who recently left OpenAI for Anthropic before resigning last week, declared that AI companies were 'gambling with our lives.' Bilal Chughtai, a former Google DeepMind researcher, stated publicly that AI 'has the potential to kill us all.' Even industry leaders have begun urging restraint — Anthropic's Dario Amodei called for a slowdown in development, a position echoed by Sam Altman, Elon Musk, and Hassabis.
Washington, however, has offered no corresponding response. President Trump has called the safety warnings 'hoaxes' and a 'conspiracy,' and his administration has shown no interest in regulating the sector. Nvidia's Jensen Huang thanked Trump for 'seeing through' the concerns, dismissing predictions of AI-caused harm as unscientific. Congress, narrowly Republican and aligned with the president, has taken no action, and approaching midterm elections make legislative movement even less likely.
The absence of government oversight has sharpened tensions within the industry itself. Aidan Gomez, CEO of Canadian AI company Cohere, accused the dominant American labs of forming a 'cartel' dressed in the language of safety. The real dispute, he argued, is not whether AI needs guardrails — that much is settled — but who writes them, who sits at the table, and whose interests the rules ultimately protect. That question remains unanswered as the technology continues to advance.
The world's largest artificial intelligence companies are moving to police themselves. On Tuesday, OpenAI announced it was joining Anthropic and Google in building a standards body designed to oversee the industry's most powerful systems before they reach the public. The model they're borrowing from is the Financial Industry Regulatory Authority, the organization that watches over American brokers and investment firms. The idea originated with Demis Hassabis of Google DeepMind, who proposed in July that such a body should test the most advanced AI systems and be funded by the industry itself to hire top technical talent and secure the computing resources needed for rigorous evaluation.
This push for self-regulation arrives as alarm about artificial intelligence's dangers has grown sharper and more public. Two former researchers at major AI labs have recently gone on record with stark warnings. Jacob Coxon, who left OpenAI to work at Anthropic, resigned last week and declared that AI companies were "gambling with our lives." On Monday, Bilal Chughtai, a former Google DeepMind researcher, amplified the concern, stating that artificial intelligence "has the potential to kill us all, and that we might be running out of time to avoid this outcome." Even prominent figures within the industry itself have begun calling for caution. On Saturday, Anthropic's CEO Dario Amodei urged a slowdown in development. Sam Altman of OpenAI, Elon Musk of xAI, and Hassabis all voiced support for that position.
Yet the political environment in Washington offers little prospect of government-mandated oversight. President Trump has dismissed the safety warnings as "hoaxes" and a "conspiracy." His administration has shown no inclination to regulate the sector. Congress, narrowly controlled by Republicans aligned with Trump, has taken no action despite the concerns. The midterm elections looming ahead make legislative movement even less likely, even as artificial intelligence and the power-hungry data centers required to run it have become issues that matter to many voters.
Trump's stance has found support in the technology industry itself. Jensen Huang, CEO of Nvidia—the chipmaker that powers the vast majority of AI systems—thanked the president for "seeing through" the warnings. Huang dismissed the prediction that AI could destroy humanity as "not grounded on science." Some leading Republicans, including Vice President JD Vance, have expressed more caution, but they remain outliers within the administration.
The vacuum left by Washington's inaction has created a different kind of tension within the AI sector. Smaller companies worry that the self-regulatory framework being built by the industry giants will entrench their dominance rather than protect the public. Aidan Gomez, CEO of the Canadian AI company Cohere, accused the dominant American labs of forming a "cartel" under the banner of safety. In a blog post on Sunday, Gomez argued that the real dispute is not whether artificial intelligence needs guardrails—that point is settled—but rather who gets to write those guardrails, who participates in the process, and whose interests the rules ultimately serve. The question of whether industry self-regulation can address genuine safety concerns while remaining fair to smaller competitors remains unresolved as the technology continues to advance.
Citations marquantes
AI companies were gambling with people's lives— Jacob Coxon, former OpenAI researcher
AI has the potential to kill us all, and we might be running out of time to avoid this outcome— Bilal Chughtai, former Google DeepMind researcher
The dispute is over who writes the guardrails, who gets to participate, and whose interests the rules protect— Aidan Gomez, CEO of Cohere
Envie de l'histoire complète ?
Lire l'original sur Deutsche Welle ↗
Technology