The US government is in advanced negotiations with AI companies to establish voluntary standards for releasing new frontier models, with an announcement possible within the next week, the Financial Times reported on Wednesday.
The framework would let officials designate a model as a "covered frontier model" and negotiate access terms before it ships, without imposing a licensing requirement the White House has explicitly ruled out. The talks build on Trump's June executive order, which asked developers to give the government early access to frontier models but stopped well short of a mandatory regime.
Washington's concern, according to the FT's sourcing, is less about consumer harm than about advanced AI ending up useful to military or intelligence services in China, Russia, or other adversarial states. Google is already in discussions, reportedly about an advanced coding model in development. Microsoft, Google, and xAI had previously agreed to pre-release government testing; Meta remains the last major holdout. OpenAI has separately been asked by the administration to slow a subsequent model's release while reviews caught up. Treasury, NSA, CISA, and NIST have until early August to build a classified benchmarking process for assessing models' cyber capabilities.
Analysis
This is Washington trying to build a regulatory structure without calling it regulation. A voluntary framework lets the White House claim oversight without defending mandatory rules in court — a political sweet spot. But the regime's force depends entirely on how many labs opt in. If Meta stays out and DeepSeek ignores it, the framework governs only the companies already most cooperative. The real test is not whether an announcement lands within the week, but whether the framework constrains anyone who was not already complying.
