U.S. policy needs to be controlling dangerous frontier AI models, says CFR's Sebastian Mallaby

Summary

Sebastian Mallaby, CFR senior fellow, argues the U.S. government is moving toward compulsory control of frontier AI models because of cybersecurity and biological weapon threats, endorsing a formal FDA-like review regime. He then analyzes the competitive landscape, saying OpenAI is being squeezed by Anthropic in enterprise and Google’s Gemini in retail, while a shift to higher expected interest rates worsens the IPO climate.

• U.S. government evolving toward compulsory control of frontier AI models to prevent cyber and bio threats
• Mallaby calls for a formal, non-arbitrary government review system similar to the FDA or FAA
• OpenAI IPO reportedly delayed to 2027 amid a tougher competitive and interest rate environment
• Anthropic seen leading in enterprise AI models, including coding and cybersecurity
• Google's Gemini model described as fast, cheap, gaining users rapidly and becoming the default on smartphones
• Chinese open-weight AI models also emerging as competitive threats
• Fed expected to hike, not cut, worsening the macro backdrop for IPOs