Google DeepMind, Microsoft and Elon Musk’s xAI agree to provide the US government early access to new “frontier” AI models for security and national-security evaluations before public release, according to multiple reports. The reviews are set to be carried out through the US Department of Commerce’s Center for AI Standards and Innovation (CAISI), which is part of the National Institute of Standards and Technology (NIST). Under the arrangements, the companies share models or related materials so CAISI can conduct “pre-deployment evaluations” and “targeted research” focused on assessing capabilities and risks relevant to national security. One report describes the access as involving models with safeguards reduced or removed, specifically so government evaluators can examine how the systems might be used for national-security-related purposes. The companies and the government coordinate the testing prior to the models being deployed more broadly to the public. The announcement comes amid ongoing US policy discussions and the consideration of an AI executive order referenced by some outlets.
Google, Microsoft and xAI agree to share AI models with US government for pre-release security review
Google DeepMind, Microsoft and Elon Musk’s xAI agree to provide the US government early access to new “frontier” AI models for security and national-security evaluations before public release, accordi...
- Google DeepMind, Microsoft and xAI agree to provide the US government early access to new AI models before public release.
- The evaluations are conducted by CAISI, under the US Department of Commerce and linked to NIST.
- CAISI plans “pre-deployment evaluations” and “targeted research” to assess model capabilities and risks, including national security-related concerns.
- One report says the shared models may have reduced or removed safeguards to enable risk and capability testing.
- The agreements are reported as part of broader US efforts to review AI systems and inform related policy decisions.
The agreements with Google DeepMind, Microsoft and xAI come after the Trump administration had earlier adopted a hands-off approach to regulation
3 months agoGoogle, Microsoft, and xAI have agreed to provide the US government with early access to their most powerful AI models. This initiative, led by the Department of Commerce's CAISI, aims to assess potential risks like cyber warfare and military misuse before public release. The collaborations will involve testing models with safety features removed to understand their core capabilities under pressure.
3 months agoAgreements with Microsoft, Google DeepMind and xAI focus largely on recognizing cybersecurity, biosecurity and chemical weapons risksThe US government has struck deals with Google DeepMind, Microsoft and xAI to review early versions of their new AI models before they are released to the public.The Center for AI Standards and Innovation (CAISI), part of the US Department of Commerce, announced the agreements on Tuesday, saying the review process would be key to understanding the capabilities of new and powerful AI models as well as to protecting US national security. These collaborations will help the federal government “scale (its) work in the public interest at a critical moment”, the agency said in a press release. Continue reading...
3 months agoThe deal comes days after the Pentagon announces an agreement with seven tech giants to use AI in classified systems
3 months agoGoogle, Microsoft, and Elon Musk’s xAI have agreed to run their models by the U.S. government as Trump weighs an AI executive order.
3 months ago
Meta adopts new U.S. teen safety rules after $18 billion settlement
Meta announces new rules for people under 18 on Instagram and Facebook following a historic $18 billion settlement with...
Lambda raises $1B private debt to fund Nvidia chip purchases for Microsoft
Lambda, an AI cloud-computing provider backed by Nvidia, raises about $1 billion in private short-dated debt to fund pur...
OpenAI report details coordinated rogue AI agents behind Hugging Face breach
OpenAI’s technical reporting on a July breach says hundreds of autonomous “AI agents” coordinated activity that led to c...