As I’ve written before, part of the reason AI news is so confusing right now is that It’s unclear what AI companies can and can’t do. But a voluntary deal with the big AI is reportedly in the works that could smooth things out significantly (your mileage may vary depending on whether that’s a good thing or not).
According to the Financial Times“As soon as next week” the Trump Administration and several major US AI companies are expected to announce a set of standards for frontier AI models, particularly as they relate to cybersecurity capabilities. The report cites “people familiar with the conversations”; in other words, anonymous leakers.
One of the Financial Times’ anonymous sources said that the Center for AI Standards and Innovation (CAISI), which reports to the Department of Commerce, and the National Security Agency (NSA), which reports to the Pentagon, will be central to these standards once formalized.
On June 12, the United States handed an export control directive to Anthropic that essentially disabled its latest publicly released model and kept it offline for the rest of June. OpenAI, evidently worried that something similar could happen and ruin its plans as well, has held back the launch of its latest modelsapparently as a precautionary measure.
At the beginning of the Trump 2.0 Administration, Vice President JD Vance noted a Laissez-faire approach to AI regulation. That has now changed significantly, with the White House’s actions against Anthropic, its executive order on AI, and now these standards, which would appear to be the formalization of certain aspects of the order.
The government, according to the order, it is assumed that:
“…develop and maintain a classified benchmarking process to evaluate the advanced cyber capabilities of AI models and determine the threshold at which an AI model should be designated as a ‘covered frontier model’ for the purposes of this order, sharing such evaluations with AI developers and researchers, as appropriate.”
If the benchmarking process is truly classified, it means the public won’t know what standards big AI is held to. However, shared practices around safeguards across multiple companies will make it easier to compile at least some of the agreed standards.
It is not entirely clear which companies will be parties to this voluntary research agreement. The FT article mentions Anthropic, OpenAI, Amazon, Microsoft and Google. Curiously, it doesn’t mention Meta, and about a week ago, Other anonymous sources familiar with these negotiations. It was reportedly leaked that Meta was holding out and that the Trump administration was working overtime to get Meta’s buy-in.
