US may unveil voluntary AI guardrails for advanced models as early as next week: Report

The United States is reportedly preparing to introduce a voluntary framework governing the release of advanced artificial intelligence models, marking another step in US' effort to strengthen oversight of rapidly evolving AI technologies.

According to a report by the Financial Times, discussions between the US government and leading AI developers have reached an advanced stage, with an announcement potentially expected as early as next week. The initiative is designed to establish common expectations for companies before they make their most capable AI systems publicly available.

While the proposed framework would not be legally binding, it is expected to outline industry-wide standards covering performance benchmarks, deployment schedules and rules on who can access frontier AI models within the US and overseas.

US steps up oversight of frontier AI

The latest discussions come as the US government increases scrutiny of cutting-edge AI models over concerns that highly capable systems could be exploited by foreign governments, including military and intelligence agencies in countries such as China and Russia.

More from Tech

The push for stronger safeguards gathered pace in June when US President Donald Trump signed an executive order directing federal agencies to work alongside leading AI developers. The order called for advanced models to undergo testing before release and instructed agencies to prepare standards governing their deployment.

As part of those broader efforts, technology companies have been engaging with government officials before launching increasingly sophisticated AI systems.

Reuters reported that Google has been holding discussions with US authorities ahead of the release of new coding-focused AI models featuring enhanced capabilities. A source familiar with the matter also told the news agency that the company has been participating in wider conversations around industry standards.

AI launches increasingly shaped by government scrutiny

The government's involvement has already influenced the rollout of several high-profile AI products in recent weeks.

On Tuesday, the US Commerce Department lifted export restrictions on Anthropic's most advanced Fable and Mythos models. The controls had been introduced less than three weeks earlier following national security concerns, temporarily limiting their availability.

OpenAI has also experienced government intervention. Last week, the company postponed the broader launch of GPT-5.6 after a request from US authorities, instead making the model available only to a limited group of vetted partners.

The heightened oversight reflects US growing focus on balancing innovation with security as AI systems become more capable and widely adopted.

The Financial Times also reported that both OpenAI and Anthropic are preparing for potential initial public offerings, making regulatory clarity increasingly important as investors and governments pay closer attention to how advanced AI technologies are developed and deployed.