Anthropic Mythos fears make Google, Microsoft and Elon Musk’s xAI make a ‘promise’ to the US government: We will not …

Google, Microsoft and Elon Musk’s xAI have entered an agreement to give the US government a “first look” at their most powerful artificial intelligence (AI) models before they are released to the public. This comes as officials in Washington are growing increasingly worried that advanced AI tools could be misused for cyberattacks or military purposes, following the release of Anthropic’s “Mythos” system, which raised alarms about AI's potential to help hackers.The Center for AI Standards and Innovation (CAISI), part of the Department of Commerce, will lead the evaluation process.“The Center for AI Standards and Innovation (CAISI) at the Department of Commerce’s National Institute of Standards and Technology announced new agreements with Google DeepMind, Microsoft and xAI. Through these expanded industry collaborations, CAISI will conduct pre-deployment evaluations and targeted research to better assess frontier AI capabilities and advance the state of AI security. These agreements build on previously announced partnerships, which have been renegotiated to reflect CAISI’s directives from the secretary of commerce and America’s AI Action Plan,” it said in a statement.According to news agency Reuters, under the agreement, the government will be able to search for risks related to cyber warfare and military misuse before tools are widely deployed, assess what these high-powered “frontier” models are capable of doing, and test versions of the software where the usual safety “locks” have been removed to see how the core technology behaves under pressure.CAISI Director Chris Fall emphasised that rigorous science is the only way to truly understand the national security implications of these new tools.“Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications. These expanded industry collaborations help us scale our work in the public interest at a critical moment,” Fall said.This deal expands on similar partnerships established in 2024 with other industry leaders like OpenAI and Anthropic. CAISI has become the central hub for this work, having already completed more than 40 evaluations of AI models – some of which have never been seen by the public.