White House issues executive order requiring voluntary AI review before launch

The White House must issue by Thursday an executive order that creates a voluntary review mechanism for new artificial intelligence models before their public launch. The measure establishes an agreement in which AI companies will share advanced versions of their systems with the government for a specified period, allowing risk assessment before making available to the market.

What does the IA executive order propose in practice?

The executive order provides for a voluntary, non-mandatory agreement between the White House and AI developers. Participants would agree to provide the government with early access to border models — those most capable and potentially dangerous — for a predefined period before public launch. During this window, federal agencies could analyze security, bias, robustness and other critical aspects.

The measure is aligned with previous management movements, such as voluntary security commitments signed with large technology companies in 2023. The difference now is formalization via executive order, which can give greater political and budgetary weight to the initiative.

Why does voluntary AI review matter to the ecosystem?

The decision reflects the growing concern with the systemic risks of increasingly powerful models. The voluntary review seeks to avoid scenarios such as the launch of systems with serious safety failures or large-scale misinformation capacity. For Brazilian developers and startups that rely on APIs or base models, the measure can signal a stricter regulatory environment in the US, indirectly influencing the global market.

In addition, the White House action takes place in parallel with other security initiatives, such as the launch of Rampart and Clarity by Microsoft, open source tools for offensive security in AI. Convergence of efforts suggests that governments and big techs are aligned on the need for more robust controls.

Technical community reaction to the executive order of AI?

Experts in AI policies received the news with optimistic caution. The voluntary approach is seen as a pragmatic first step, but critical point out that without enforcement mechanisms sufficient adherence may be lacking. Companies like OpenAI, Google and Anthropic have already demonstrated public support for security principles, but practical implementation still depends on details of the executive order, which should be disclosed after signing.

The information was originally disclosed by CNN, which highlighted the urgent nature of the measure and the expectation of announcement already on Thursday.

Frequently Asked Questions about the executive order of AI

Is the executive order mandatory for all AI companies?

Nope. The deal is voluntary, but the White House expects large developers to sign up voluntarily to demonstrate commitment to security.

What kind of models will be reviewed?

Border models — the most advanced and able to cause systemic damage — will be prioritized in sharing with the government.

When does the executive order come into effect?

The signature is scheduled for Thursday, but the implementation schedule and operational details will be released after the official announcement.