Spooked by Mythos, Trump suddenly realized AI safety testing might be good
Summary
Ars Technica reports that the Trump administration has shifted to support AI safety testing, signing government testing agreements with Google DeepMind, Microsoft, and xAI to evaluate frontier AI models before and after release. The piece analyzes CAISI’s role, the lack of defined testing standards, concerns about funding and independence, and the risk of politicization in AI governance, with expert voices weighing potential paths forward such as independent audits.