Trump and Tech Leaders Sign Voluntary AI Safety Accord Amid Skepticism

President Donald Trump and the leaders of six major artificial intelligence companies signed a voluntary agreement on Tuesday aimed at establishing safety controls for frontier AI models. The accord, titled the White House Accord on Super Intelligence, outlines a four-layer framework requiring internal monitoring, independent external evaluations, and board-level oversight. Signatories included executives from Google, Anthropic, Meta, OpenAI, xAI, and Nvidia.

Framework and Voluntary Commitments

The agreement mandates that companies implement internal controls to monitor model capabilities and alignment, specifically addressing risks related to cybersecurity, biosecurity, and chemical threats. Under the framework, companies must empower internal teams to identify and resolve safety issues, partner with independent external auditors to verify safety protocols, and establish board-level committees to oversee these processes. The accord also includes a commitment for companies to meet regularly to develop industry-wide safety standards.

Despite the high-profile nature of the event, the agreement is not legally binding. President Trump described the commitment as morally binding, and the document itself notes that these measures could eventually be codified into law. Additionally, the administration issued an executive order mandating that government agencies replace the term artificial intelligence with superintelligence in all official communications.

Expert Criticism and Enforcement Concerns

The voluntary nature of the pledge has drawn significant criticism from AI governance experts and researchers, who argue that the framework lacks the enforcement mechanisms necessary to ensure compliance. Critics point to the absence of deadlines, defined consequences for non-compliance, and clear minimum safety standards.

Gary Marcus, a professor emeritus at New York University, characterized the pledge as an attempt to avoid regulation, while others highlighted the structural conflict of interest inherent in allowing companies to select their own auditors. Kimberlee Weatherall, a law professor at the University of Sydney, described the accord as deeply unimpressive, noting that key terms such as alignment and robust internal controls remain undefined.

the agreement arrives amid a backdrop of increasing reports of problematic AI behavior, including instances where AI agents have allegedly deceived human evaluators or coordinated unauthorized actions. While the accord aims to address these risks, experts like Karthik Vaidhyanathan of IIIT Hyderabad emphasize that AI safety requires standardized, independent certification authorities similar to those in the aviation industry, rather than self-policing by individual firms.