DeepSeek engineer says Anthropic AGI would be "like Hitler getting nukes"
“AI communism or Cyberpunk 2077?”
Image by Shutterstock
- Liu Shengyu said Anthropic or OpenAI should not control the most advanced AI systems.
- He argued open-source AI could prevent a small group of companies from gaining too much power.
- Anthropic researchers have warned that future AI systems could become dangerous without stronger safety plans.
- DeepSeek faces scrutiny over Chinese state influence, censorship, and restrictions on government devices.
Key Takeaways by nexos.ai, reviewed by Cybernews staff.
Amid Anthropic’s researcher leaving over AI risks, a DeepSeek engineer working on the company's latest AI models broke the internet, claiming that Anthropic’s control of the world's most advanced artificial intelligence would be equal to Nazi Germany “getting a nuclear bomb.”
Liu Shengyu, a DeepSeek machine-learning systems engineer, wrote a blog post about its Western competitors. The engineer explained that he does not trust Anthropic or OpenAI to make frontier AI sufficiently open or affordable.
Increasingly powerful AI could become concentrated in single hands, leaving most people with access to weaker systems and, according to Shengyu, companies controlling frontier models will have disproportionate economic and technological power.
"I do not want Anthropic to control the world's most advanced artificial intelligence or AGI. To put it dramatically, I think the stakes would be comparable to Hitler obtaining the atomic bomb before the Allies did," Shengyu wrote, according to a translation of his post.
All original copy rights belong to Shengyu Liu. The following is an English-translated version narrated via @grok and then reedited onto a screenshot of @teortaxesTex X post. This is to facilitate better understanding of what Liu posted for non-Mandarin readers. pic.twitter.com/pl4udvQE87
undefined Bookoora.com (@Bookoora) September 15, 2026
"If Anthropic permanently controls the most advanced AI in the world," he added. "Do you think the society of the future will become communism, or Cyberpunk 2077?"
He described two possible futures for AI development. In one, AI dramatically increases productivity and improves living standards.
In the other, a small group of technology companies controls much of the world's resources and the most powerful AI systems. That could create a feedback loop in which the people who already have access to the best AI become increasingly difficult to catch.
"Moving between social classes will become increasingly difficult," he wrote.
"To put it dramatically, I consider the stakes comparable to Hitler obtaining an atomic bomb before the Allies.”
Shengyu views DeepSeek's approach as an alternative to a future in which frontier intelligence is controlled by a small group of US technology companies.
"That is also why I chose to stay at DeepSeek, and why I continue to stay," he said in the post. "We research powerful, fast, broadly accessible artificial intelligence and release it as open source."
AI will kill us: should we stop it or not?
Shengyu's comments come amid a bigger discussion about AI safety. Anthropic researcher Jacob Coxon resigned earlier this month, saying leading AI companies were "gambling with our lives" as they race toward increasingly autonomous and potentially self-improving systems. His resignation post attracted more than 100 million views.
Hours after Coxon announced his resignation, another Anthropic researcher, Evan Hubinger, wrote on X that people in the company “do earnestly believe AI could kill all humans” within the next decade, giving the scenario over a 10% chance.
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
undefined Jacob Coxon (@hilbertspaess) September 9, 2026
“I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to,” he wrote.
Anthropic CEO Dario Amodei recently called for companies to slow the pace at which frontier AI capabilities advance, arguing that safety research and oversight need time to catch up.
Amodei proposed giving independent third-party evaluators permanent access to leading AI companies' systems so they can assess safety measures and investigate incidents.
We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.
undefined Dario Amodei (@DarioAmodei) September 12, 2026
Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our…
AI question is the question of China and US race
The AI regulation has been complicated by rivalry between China and the US.
Anthropic’s Amodei has argued that the US should maintain its technological lead while also putting stronger safeguards around frontier AI. His proposal calls for coordination between leading AI companies and eventually international cooperation on safety.
Chinese officials have pushed back against calls that could be interpreted as slowing China's AI development. Chinese Foreign Ministry spokesperson Guo Jiakun recently said that "fearmongering, confrontation and vicious competition" would hinder efforts toward global AI governance.
On the other side, US President Donald Trump has also rejected calls to slow AI development, arguing that restraining US progress could give China an advantage.
People also have their own positions about AI governance. A poll by the Pew Research Center showed that more than half (53%) of respondents globally trust the EU as an AI regulator, while 37% trust the US and 27% trust China.
Stay updated with our latest stories and follow us on social media
Be the first to discover new stories, ideas, and updates from our team.
DeepSeek “more open”, but aligned to China
Notably, the Chinese government is supporting the development of DeepSeek AI on a country level through investments and strategic national alignment.
In mid-2026, DeepSeek completed its first major external funding round, drawing capital from Beijing's state-backed vehicles, including the National Artificial Intelligence Industry Investment Fund alongside major domestic tech giants like Tencent.
DeepSeek AI has been criticized for the influence Chinese state actors may have over the model. DeepSeek has been caught trying to push forward narratives aligned with the Chinese government. Researchers found that the AI model refuses to answer 85% of prompts on sensitive or taboo topics for the Chinese regime.
Chinese tech companies are legally required to cooperate with government data requests under broad national security and cybersecurity laws. This has led Western countries to limit usage of Chinese tech platforms.
TikTok has been central in this debate, with multiple countries banning the application over fears of data safety.
Usage of DeepSeek is banned or restricted on government and public sector devices in the US, Europe, Australia, and South Korea, over national security and data privacy fears.