Google's DeepMind engineer defends company's AI capabilities amid claims of falling behind

On September 10, 2026, AI researcher Ethan Mollick claimed on X that Google no longer possesses cutting-edge models, potentially jeopardizing its position in mathematics and cybersecurity. In response, Logan Kilpatrick, a DeepMind engineer, asserted that Gemini 3.8 Cyber outperforms Anthropic's Mythos in cybersecurity tests and real-world applications. The model is already in use by Google's Chrome, Wiz, and Cloud teams. Kilpatrick emphasized the focus on empowering cybersecurity professionals and highlighted that the Gemini 4 series will continue advancing the field. The debate underscores ongoing competition in AI development, with claims of superiority driving innovation in critical domains like cybersecurity.