2 More Google Gemini Engineers Defect to Anthropic, Stock Dips 1.2%

For the AI research community, these defections signal a shifting talent landscape where pre-IPO equity at Anthropic and OpenAI is overshadowing even Google’s vast resources. The loss of Adler, a specialist in AI coding, and Pritzel, who helped train Gemini’s core systems, could directly delay upcoming model iterations and weaken Google’s position in critical areas like code generation and safe AI. As startups dangle IPO windfalls, the industry is watching whether Google’s centralized research structure can still retain the innovators who define the frontier.

Google's artificial intelligence division suffered another significant talent blow as two senior researchers, Jonas Adler and Alexander Pritzel, confirmed plans to join rival Anthropic. The departures, first reported on June 24, 2026, mark the latest in a growing brain drain that has already seen Nobel laureate John Jumper and star researcher Noam Shazeer exit in recent days, while senior research engineer Arthur Conmy also announced his move to Anthropic via social media. Adler, who worked on AI coding efforts, and Pritzel, involved in training systems for the Gemini model, were considered key internal contributors to Google's flagship AI platform. Their loss raises serious questions about the search giant's ability to maintain momentum in the fiercely competitive AI landscape, where it has only recently begun to close the gap with Anthropic and OpenAI after a slow start. Investors reacted negatively, with Alphabet shares closing down slightly after an intraday drop of up to 1.2%, underscoring market sensitivity to Google's talent retention challenges.

“

Google's artificial intelligence division suffered another significant talent blow as two senior researchers, Jonas Adler and Alexander Pritzel, confirmed plans to join rival Anthropic.

The exodus is not occurring in a vacuum. Anthropic and OpenAI are both on the cusp of initial public offerings, offering pre-IPO equity packages that are uniquely attractive to top-tier researchers who could earn windfalls far exceeding what Google's established compensation structures can provide. This financial incentive, combined with the allure of joining fast-moving, AI-native startups, is proving difficult for Google to counter. Internal dynamics are also at play: sources indicate that Shazeer's decision to leave for OpenAI was influenced by a reallocation of critical computing resources from his project to a London-based team at DeepMind, a move intended to streamline efforts but that ended up alienating key talent. Such resource battles highlight a broader tension within Google—the struggle to balance legacy-scale infrastructure with the agility demanded by cutting-edge AI research.

These defections threaten to undercut Google's Gemini platform at a critical juncture. Adler's expertise in AI coding and Pritzel's work on training methodologies are directly aligned with the type of capabilities that differentiate leading models. Anthropic's Claude has already established itself as a formidable competitor, and the addition of Adler, Pritzel, Jumper, and Conmy could accelerate its development in code generation and safety-focused AI—areas where Google hoped to press its advantage. For OpenAI, the acquisition of Shazeer, known for his work on transformative language models, strengthens an already deep research bench and signals that the startup remains a magnet for top-tier talent.