Trump Administration Finalises New AI Model Evaluation Framework

The Trump administration has finalised a planned voluntary framework for evaluating new artificial intelligence (AI) models. A White House official confirmed that a meeting with industry partners will be held on Tuesday (4 August) to discuss the details of its implementation. Although the official did not provide specific details regarding the content of the framework, the move follows an executive order signed by President Trump in June. The directive aims to enhance AI safety and innovation by establishing a programme for technology companies to voluntarily share powerful new models with the government before their public release. The executive order stressed that the federal government does not wish to hinder innovation with overly burdensome regulations. Instead, the policy is designed to strengthen the national cybersecurity system through the utilisation of AI technology. Within this framework, there is a process to identify so-called frontier models—AI systems at the cutting edge of the field. The government will work with companies willing to provide federal access to review these models for up to 30 days prior to their official launch. This step is taken amidst concerns that the capabilities of advanced AI models could be used for malicious purposes, such as identifying previously overlooked software vulnerabilities in critical systems. For instance, the AI laboratory Anthropic announced in April that it provided its new model, Mythos, to selected partners to strengthen defences against cyberattacks before the technology was made widely available. The White House reiterated that collaboration between AI companies and the government is voluntary. The policy is claimed not to prohibit AI innovators from continuing to advance their technology, but rather to create a safer and more measured development environment.