US Order on AI Oversight: What It Means for AI Governance

The United States has issued a new executive order that asks makers of the most advanced artificial intelligence (AI) systems to let the government check their models before they are released to the public. The order was signed on June 3, 2026. It is called 'Promoting Advanced Artificial Intelligence Innovation and Security'. This is a notable shift, because the US government had earlier preferred a light-touch, hands-off approach to AI.

What the order does

The order sets up a voluntary framework. Under it, federal agencies will decide which systems count as "covered frontier models" — meaning the most powerful and capable AI models. Companies that build such models can choose to take part. If they agree, they will give certain government agencies early access to test the model for security before it is launched widely.

The main goal stated by the government is cybersecurity. As AI systems become more powerful, they may be able to find or misuse weaknesses in computer systems. Early testing is meant to help protect critical infrastructure — such as power, banking, and defence networks — from such risks.

The 30-day review window

A key feature is a 30-day review period. Participating developers would allow chosen government agencies to test an advanced model for up to one month before public release. Agencies expected to be involved include the Departments of Defense, Commerce, Homeland Security, and Treasury.

An earlier draft of the order had proposed a 90-day (three-month) review. That version was held back after the technology industry warned that a long review could slow down innovation and weaken US companies in the global race, especially against China. The final order cut the review to 30 days so that companies can fit the testing into their normal pre-launch work without major delays.

Why "voluntary" matters

The order is deliberately voluntary. It does not create any new licence requirement and does not force companies to get approval before launching an AI model. Instead, agencies are asked to design a cooperative framework with developers. The order also calls for stronger cyber defences across government systems and closer coordination with industry to find software weaknesses.

What is AI governance?

AI governance refers to the rules, policies, and institutions that guide how AI is built and used. The aim is to encourage useful innovation while reducing risks such as security threats, misuse, bias, and harm to citizens. Governments around the world are trying to find the right balance: too little oversight can allow dangerous misuse, while too much can slow progress and push companies abroad. The US order is one example of this balancing act — it keeps an "innovation-first" stance while trying to manage new cyber risks from powerful AI tools.

The India angle

India has also been thinking about how to regulate AI. On March 1, 2024, India's IT Ministry issued an advisory that asked AI platforms to seek government permission before launching services that were still "under-testing" or "unreliable". The move drew strong criticism from the technology industry, and the government revoked the requirement within about a fortnight. India later clarified that the advisory was meant for large, significant platforms and not for startups.

India's broader effort to build AI capacity is led by the IndiaAI Mission, a government programme that supports computing power, datasets, skilling, startups, and safe and trusted AI. The country has generally favoured an approach that promotes innovation while building safeguards, rather than imposing heavy pre-launch approvals.

The bigger picture

The US order shows a wider global trend: countries are moving from no rules at all toward at least some structured oversight of the most powerful AI systems, often focused on security. Whether such steps will meaningfully reduce real-world cyber risks is still debated, but the direction — more attention to AI safety and governance — is clear.