OpenAI and Anthropic Are Quietly Teaming Up in Washington

OpenAI and Anthropic are arch enemies whose leaders can’t stand each other. But the two foes have formed an unholy alliance in Washington in the interest of self-preservation. 

With the Trump administration set to finalize its framework for regulating and evaluating “frontier models” by August 1, the two leading AI developers are quietly lobbying D.C. on a shared agenda: making sure lagging competitors like Meta and SpaceXAI also have to comply with government reviews of their models, and raising alarms about Chinese open-source models.

Their alignment isn’t obvious at first glance. But our sources tell us Anthropic and OpenAI have pushed to expand the type of models that would require government review under the framework to include those from their competitors. (They’ve discussed how the government and industry can develop an evaluation to determine whether a model requires extended review, based on its capabilities in cybersecurity or other tasks that could present national security risks.)

Some of those rivals have disagreed, preferring that the government focus on Anthropic and OpenAI’s models, given that those models are considered to be the most capable (and thus potentially dangerous and in need of a review).