Chinese AI Star Walks Away from Anthropic After ‘Adversarial Nation’ Remark, Joins Google DeepMind
The world of artificial intelligence just lost one of its brightest young minds to a rival. Yao Shunyu, a 28-year-old researcher whose work helped shape the latest generation of language models, has left Anthropic after less than a year and jumped to Google DeepMind. His reason? A single phrase: Anthropic’s decision to label China an “adversarial nation.”
Yao didn’t mince words in a candid blog post this week. While he said many colleagues inside Anthropic probably disagreed with the wording, he couldn’t remain at a company that cast his home country in that light. “I don’t think there is a way for me to stay,” he wrote in a farewell essay explaining Why Anthropic, and why leaving?
A meteoric rise from Beijing to Silicon Valley
Yao’s story reads like a case study in the global AI talent pipeline. After graduating from Beijing’s prestigious Tsinghua University—where he earned top honors including the Yeh Chi-Sun prize—he crossed the Pacific for a PhD in physics at Stanford. Postdoctoral work at UC Berkeley polished his skills before he shifted gears into AI. By October 2024, he was at Anthropic, diving straight into research that fed into Claude 3.7 Sonnet, released just months later.
The excitement, he admitted, was intoxicating. “After being a physicist for many years, it was so exciting to see your research impact the frontier model capability immediately,” he recalled. He even described Anthropic as one of the best places for physicists and other STEM PhDs to start a career in AI. But the glow didn’t last.
The breaking point
Last month, Anthropic tightened access to its technology, banning subsidiaries of Chinese companies worldwide and explicitly calling China “adversarial.” The move reflected growing U.S. fears about intellectual property theft and the military use of advanced AI. For Yao, that decision accounted for “about 40 percent” of why he left. The other 60 percent, he hinted, stemmed from internal issues he declined to share publicly. His parting words carried both finality and sting: “It was good with you, but it is better without you.”
A big win for DeepMind
Now Yao heads to Google DeepMind, where he’ll work on the Gemini team, the engine behind Google’s most ambitious multimodal models. DeepMind’s chief executive Demis Hassabis has tried to strike a more conciliatory tone on geopolitics. Earlier this year, he urged Washington and Beijing to cooperate on AI safety, warning that the risks of advanced systems cut across borders. That message stands in stark contrast to the more hawkish approach of Anthropic’s leadership.
Amodei’s long shadow
Anthropic CEO Dario Amodei has been outspoken about China’s role in AI, echoing arguments once common in Washington under the Trump administration. He’s warned against shipping powerful chips to Chinese firms, saying it would “mortgage our future” by advancing military tech. His position isn’t abstract—his early career included a stint at Baidu’s Silicon Valley AI Lab, where he worked on speech recognition project Deep Speech 2 with LSTMs. According to our exclusive source from Baidu, Amodei’s frustrations with Baidu’s rigid management style and focus on “obedience” over open inquiry shaped his hardline stance today.
Amodei insists his concerns target state-driven risks, not individuals or diaspora researchers. “It’s about strategic risks from specific firms and actors, not people,” he said recently. Yet critics argue the rhetoric creates an atmosphere of exclusion. Yao’s very public exit now serves as a case study for that claim.
A talent race with no borders
Yao’s move underscores the bigger question hanging over the AI arms race: How do you protect national security without pushing away the very minds driving progress? A 2022 study showed that nearly four in ten top AI researchers in the U.S. were born in China, outpacing even American-born experts. If the climate turns hostile, that talent could scatter to other labs or other countries.
For Yao, though, the decision appears personal. He’s a physicist-turned-AI researcher chasing the next frontier, and DeepMind now has him on board. His journey highlights a truth that every company in this space must face: innovation doesn’t recognize passports, but policies and perceptions can still shape where the brightest minds choose to work.