Something unusual happened this week. The CEOs of Anthropic, OpenAI, and xAI — three companies racing each other to build ever more powerful AI — stood together and asked for the brakes to be applied. And the US President said no.

On September 12, Anthropic CEO Dario Amodei published “We Must Pace the Frontier,” a roughly 3,800-word essay warning that without deliberate action, swarms of rogue AI agents could threaten large parts of the internet within six to twelve months. OpenAI’s Sam Altman and Elon Musk both said, within a day, that they agreed with him — an extraordinary moment of consensus from three executives who compete fiercely for talent, customers, and compute.

By September 13, President Trump had a different answer.

“We’re leading China in AI. We’re the most sophisticated country in the world, and frankly, I want to keep it that way — because whoever wins AI, wins,” Trump told reporters. He dismissed the dire warnings as exaggerated, saying “negative forces” were making claims about things that would not happen. Speaker of the House Mike Johnson echoed the position from Congress: AI safety regulation would not come from Capitol Hill. The responsibility, he said, falls on the companies themselves.

And so here we are.

The Regulatory Vacuum

What this week’s events reveal is less a debate about risk and more a structural fact about where AI governance will — and won’t — come from in the near term.

The federal government is not going to regulate frontier AI. That is now clearly stated policy. Congress is not going to lead. The executive branch sees AI as an arms race against China where pulling back is losing.

At the same time, the three biggest AI labs have been quietly holding working-group meetings since July, working toward an industry-led standards body that would share protocols for testing models before release, conduct independent pre-release evaluations, and create standardized risk assessments. Amodei has been the driving force; Alphabet’s Demis Hassabis has reportedly pushed to have the organization established before the end of 2026.

This is meaningful — but it is voluntary. There is no enforcement mechanism. Companies that do not sign up are not penalised. And the very same labs proposing these standards have spent the past two years racing to ship the most capable models they can.

What This Means for Your Business

If you are running a business that is adopting AI, or planning to, the government is not going to hand you a compliance framework. What fills that gap will be a patchwork of industry standards, enterprise vendor policies, and your own internal governance decisions.

A few things worth understanding:

The EU is moving differently. While the US steps back, the EU’s AI Act is moving from preparation into active enforcement. September 2026 marks the shift from administrative groundwork to real statutory compliance deadlines. If your business touches European customers or data, you have a different regulatory reality than a pure US operation.

Your AI vendors are writing the rules. In the absence of government standards, OpenAI, Anthropic, Microsoft, Google, and Salesforce are effectively setting the norms through their usage policies, safety commitments, and enterprise agreements. Understanding what your vendor commitments actually include — and don’t include — matters more now.

The risk of inaction is not regulatory. The business risk of poor AI governance is not a fine from Washington. It is operational failure, data exposure, reputation damage, or deploying agents that do things you did not authorise. Those risks are real regardless of what the government does.

Industry self-regulation will create winners and losers. If an industry-led safety body does emerge and establishes credible pre-release testing standards, companies that adopt them early will be better positioned with enterprise customers who are demanding vendor accountability. Being ahead of informal standards matters even without a formal law.

The Bigger Picture

Amodei’s essay and Trump’s rejection reveal a genuine philosophical divide about what AI represents right now. Amodei sees a technology approaching a threshold where the risks shift from theoretical to operational. Trump sees a geopolitical competition where caution is the same as ceding ground.

Both perspectives contain real things. The risk concerns are not invented — the same researchers building these systems are raising them. The competitive pressure is also real — the US does lead China in frontier AI, and the gap could close.

For business leaders, the honest takeaway is that neither the government nor the AI labs will tell you exactly what to do with AI inside your organisation. You have to build that judgment yourself: what your systems are authorised to do, what human oversight is required, where your data goes, and what you do when something goes wrong.

That is harder than waiting for a regulation to comply with. But it is also an opportunity. The companies that build thoughtful AI governance frameworks now — not because they have to, but because they choose to — will be better prepared for whenever the rules do come. And they will come.

---

Enterprise DNA helps business leaders build practical AI adoption strategies. If you want to understand what responsible AI governance looks like in a growing business, book a call with our advisory team.