US Cedes Its AI Governance Responsibilities to Others

•
July 20, 2026    

The economic strength of the United States, its heritage of entrepreneurial innovation and leading educational institutions have been a magnet for the world’s tech talent. U.S. tech firms are once again at the forefront, with a handful of companies leading the rise of artificial intelligence.

See Also: Moving Past Legacy AD Bridging With Idira Identity Bridge

But despite these advantages, America now appears to be ceding its leadership role in managing the risks of AI to Silicon Valley and other nations – including China.

It’s widely acknowledged that the rapid progress of AI tools, and the awesome potential for future developments, represent equally vast opportunities and existential challenges to humanity. While the technical world grapples with safeguards, guiderails and harnesses, the debate has shifted from “responsible AI” to nationalistic power, incentives and legitimacy. The big question is, who’s in control? Who sets the parameters, for whose benefit and within what cultural or ethical paradigm?

Governance is becoming central to the development of AI. But so far in the United States, that power has been ceded to the corporations.

Frontier AI companies are deciding who can access their advanced models, with Anthropic’s Glasswing and OpenAI’s Daybreak programs inviting only favored customers to use their latest AI models. The U.S. government has only stepped in requiring frontier firms to share the new tools for a voluntary preview, but no one knows what the government is evaluating and whether it even has the authority to regulate these models. This inconsistent approach to AI governance, informed by an underlying antipathy to regulation, has opened the door for others to set the pace.

Early this month, Demis Hassabis, CEO of Google DeepMind, posted a 1,500-word article on X titled “A Framework for Frontier AI and the Dawning of a New Age.” It calls for a U.S.-led, federally overseen AI standards body with mandatory pre-deployment testing, higher safety obligations for frontier labs and an international framework to slow development if risks become “unacceptable.”

So far, the U.S. government hasn’t responded to this call to action. Meanwhile, China on July 16 created its own international organization called the World Artificial Intelligence Cooperation Organization, or WAICO, “for the positive, for good and for humanity.” WAICO comprises 29 countries including China, Russia and Brazil – but not the United States. Meanwhile, U.S. allies in the European Union will start enforcing key provisions of the EU AI Act on Aug. 2. The United Kingdom’s evolving AI safety agenda and various initiatives in other countries all reflect a growing recognition that frontier AI is not just another product category. It comes with major threats related to cybersecurity and national security that require mandatory controls.

While the U.S. government lags behind, the private sector is taking its own path to AI safety. For example, Anthropic is hiring dozens of highly paid experts to prevent its models from being used to design explosives, build nuclear weapons, run cybercrime operations or conduct large scale financial fraud.

Experts argue even though the Trump administration could issue somewhat capricious export controls on Anthropic’s Mythos and request a preview of OpenAI’s GPT 5.6, the principle of government oversight remains preferable to leaving frontier AI governance entirely in corporate hands.

As Hassabis noted, “Funding would need to be substantial and likely mostly come from industry, in order to attract world-class technical talent and provide the necessary compute resources for large-scale testing.” Anthropic’s latest hiring spree offers salaries exceeding $200,000, with at least one job topping $400,000, demonstrating what it takes to attract world-class talent.

But one crucial point remains: Any serious governance framework has to be international. Frontier AI is a truly global concern. Models are trained on data from everywhere, deployed across borders and used by actors whose incentives may be misaligned with any single country’s interests.

If the benefits of AI are to serve humanity as a whole – rather than just billionaire tech bros, countries at the leading edge or, even worse, countries with dictatorial control over their populace – then the rules that shape its development and use can’t be purely national or corporate.

Hassabis’ proposal and Anthropic’s fear that AI could be used to build weapons of mass destruction should be sending a strong message to governments worldwide: Frontier AI needs oversight that’s technically competent, properly resourced and structurally aligned with the public interest.

Tech industry money will invariably be part of the picture, but the authority to decide AI safety controls, access controls and when to slow down launches shouldn’t rest solely with the AI labs themselves. It’s time for the United States to take an enlightened leadership role, or risk losing it to the self-serving interests of others.

Yes, we’re at the dawn of a new age. The question is not just how powerful will frontier AI become, but who gets to decide how that power will be used?