On May 11, an advocacy group, Americans for Responsible Innovation, urged the Trump administration to implement mandatory safety reviews for artificial intelligence models before their public release, specifically targeting those with potential cyberattack or weapons development capabilities. The group emphasized that companies that fail such reviews should be ineligible for lucrative U.S. government contracts, particularly those spending over $100 million annually on computing resources or generating at least $500 million in AI revenue. Currently, the U.S. Center for AI Standards and Innovation conducts voluntary reviews for models developed by major players like OpenAI and Anthropic, but the group advocates for more stringent, enforced requirements similar to California’s existing safety reporting regulations.

xAI: xAI develops frontier AI models aimed at advancing scientific discovery. It recently joined voluntary safety review agreements with the U.S. Center for AI Standards and Innovation. The company’s models fall under proposed thresholds for mandatory government screening.
Google: Google develops advanced AI technologies and participates in U.S. government AI safety initiatives. It recently joined voluntary agreements for model reviews with the U.S. Center for AI Standards and Innovation. The company’s involvement supports broader efforts to assess risks in frontier AI systems.
OpenAI: OpenAI creates frontier AI models and has recently secured Pentagon contracts by accepting safety boundaries aligned with U.S. law. It shares pre- and post-deployment models with the U.S. AI Safety Institute under a formal agreement. The company is involved in voluntary AI safety evaluations by government bodies.
Anthropic: Anthropic develops advanced AI models with a focus on safety through its Responsible Scaling Policy. Its model Mythos has raised concerns about enabling complex cyberattacks, prompting calls for pre-release government screening. The company participates in voluntary safety reviews with the U.S. Center for AI Standards and Innovation.
Microsoft: Microsoft builds AI models and infrastructure, collaborating with U.S. agencies on safety protocols. It has entered voluntary agreements allowing the U.S. Center for AI Standards and Innovation to review its AI systems. This participation aligns with calls for standardized safety vetting of frontier models.
U.S. Government: The U.S. Government oversees AI safety through agencies like the Department of Commerce’s National Institute of Standards and Technology. It has established voluntary agreements with leading AI companies for model testing and evaluation. The Trump administration is being urged to implement mandatory safety reviews for frontier AI models to mitigate national security risks before public release.
Americans for Responsible Innovation: Americans for Responsible Innovation is an advocacy group pushing for responsible AI development. It sent a letter to Trump administration officials urging mandatory safety reviews for frontier AI models. The group recommends tying government contract eligibility to passing these reviews and establishing a permanent enforcement office.
U.S. Center for AI Standards and Innovation: The U.S. Center for AI Standards and Innovation, also known as US AISI or CAISI, leads government efforts in AI safety research, testing, and evaluation. It conducts voluntary reviews of AI models from companies including OpenAI, Anthropic, Google, Microsoft, and xAI. Advocacy groups propose it develop mandatory requirements for frontier models.

Safety Advocacy: Americans for Responsible Innovation urges mandatory pre-release safety reviews for AI models capable of cyberattack or weapons development risks.
Voluntary Framework: U.S. Center for AI Standards and Innovation reviews models from OpenAI, Anthropic, Google, Microsoft, and xAI through non-binding agreements.
Contract Eligibility: Proposed rules would bar AI companies failing safety reviews from U.S. government contracts, targeting those with high compute spending or AI revenue.