US AI companies will be asked by the Trump administration to participate in voluntary security assessments.

The Trump administration has issued an executive order, requiring agencies to establish partnerships with AI developers for voluntary security evaluations. This development signals a change of approach from the president's earlier stance on regulating technology.

Leading US AI companies are set to receive a request from the Trump administration to participate in voluntary security evaluations of their most advanced models prior to public release, as mandated by a recent executive order issued on Tuesday amid growing concerns about AI system vulnerabilities.

President Donald Trump has issued an executive order instructing key government departments, including Treasury, Defense, Commerce, and Homeland Security, to collaborate with AI developers on voluntary security evaluations.

The Trump administration's order stipulates that U.S. agencies have a 30-day window to evaluate AI models prior to their release to external entities. Concurrently, the order prioritizes enhancing cybersecurity across all government departments.

Trump's stance on AI is undergoing a significant transformation as evidenced by this latest directive. His administration had previously advocated for minimal government intervention in the tech industry, urging states to refrain from implementing regulations he deemed restrictive. This new approach indicates a more engaged federal role in overseeing AI advancements.

Implementing voluntary testing for the auto industry may lead to financial losses if it delays the introduction of new models, forcing manufacturers to adapt their performance to meet heightened security standards.

Alphabet's GOOGL.O subsidiary, along with Anthropic and OpenAI, initiated discussions with the US government regarding cybersecurity in May. A high-ranking US official disclosed this information during a conference call with journalists.

Kent Walker, a Google executive, hailed the executive order as a significant advancement, whereas Anthropic expressed eagerness to collaborate with the White House on executing this directive.

Sam Altman, OpenAI's CEO, believes the directive strikes a fair equilibrium. He advocates for the US to spearhead AI advancements, cultivating top-tier models while ensuring their safety and equipping trusted cybersecurity experts with cutting-edge tools.

A planned executive order on artificial intelligence was put on hold by Trump on May 21, as he expressed dissatisfaction with specific provisions. He opted against proceeding, citing concerns that certain actions could compromise US interests in the ongoing AI rivalry with China and undermine its competitive edge.

Anthropic, OpenAI and Alphabet's along with Anthropic and OpenAI, initiated discussions with the US government regarding cybersecurity in May. A high-ranking US official disclosed this information during a conference call with journalists.

Developing the order involved consultation with banks by the Treasury Department, according to Scott Bessent, who serves as Treasury Secretary. In mid-May, Bessent emphasized the importance of collaboration with AI developers and critical infrastructure providers. Together, they will identify software vulnerabilities through scanning and create patches to address these weaknesses. The focus is on sectors that are crucial to the US economy and require enhanced protection against potential threats, including banks, emergency services and healthcare facilities.

The US Department of Commerce's Center for AI Standards and Innovation has been evaluating AI models submitted by companies like OpenAI and Anthropic for several years now. This voluntary federal testing process involves a rigorous scrutiny of these models to ensure their security and compliance with standards. In May, the department made headlines when Google, xAI, and Microsoft agreed to participate in this evaluation, although the details of their involvement were later removed from the website.