Google and AI researcher argue over frontier model status

AI researcher Ethan Mollick posted on X that Google no longer has a frontier model, saying Google will fall behind in math and cybersecurity. Google DeepMind engineer Logan Kilpatrick pushed back, saying Gemini 3.8 Cyber has already surpassed Anthropic's Mythos in many cybersecurity tests and real-world capabilities.

Kilpatrick said Google's internal Chrome, Wiz and Cloud teams already use the model widely. He stressed that the team's focus is giving more cyber defenders stronger capabilities.

He added that the Gemini 4 series will keep pushing frontier cybersecurity models forward.