趋势解读：MAGA-aligned groups want government oversight of frontier AI，提升开发者接入体验

A conservative coalition led by Humans First is calling on President Trump to mandate safety testing for frontier AI models before release. Prominent signatories, including Trump's former chief strategist Stephen K. Bannon, warn of serious risks to elections, cybersecurity, and critical infrastructure. The group doesn't trust tech companies to police themselves and wants government audits similar to nuclear and aviation oversight. A coalition of conservative organizations led by Humans First has called on President Donald Trump in an open letter to issue an executive order requiring mandatory safety testing for frontier AI models before they ship. Signatories include Amy Kremer, chair of Humans First and longtime Tea Party activist, Stephen K. Bannon, Trump's former chief strategist and host of the right-wing War Room podcast, Brendan Steinhauser from the AI-critical Alliance for Secure AI, and Jason Van Beek from the Future of Life Institute, which has pushed for AI regulation for years. The list also features Michael Toscano from the conservative Institute for Family Studies, Ryan Girdusky from the right-wing 1776 Project PAC, and psychologist Geoffrey Miller from the University of New Mexico. Dozens of pastors and regional activists from the MAGA orbit round out the group. Ad The group warns that powerful AI systems could pose threats to cybersecurity, critical infrastructure, financial systems, election integrity, biosecurity, and the military. AI is already being used for sophisticated cyberattacks, targeted disinformation, and automated fraud, they argue. The next generation of models carries risks on a different scale entirely. Ad DEC_D_Incontent-1 Private companies shouldn't be allowed to deploy such technology without government review, the letter states. Some firms have openly discussed selling their most capable systems to countries with different values. The companies can't be trusted to police themselves. The authors draw parallels to existing controls on nuclear material and aviation certification. Powerful AI models that could help design bioweapons or break into critical infrastructure deserve the same level of scrutiny, they argue. Ad Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.