DeepSeek engineer attacks Anthropic, OpenAI over AI 'pacing' push
Synopsis
A DeepSeek engineer behind the V4.1 models publicly declared he does not want Anthropic to control AGI, invoking Nazi Germany to warn against AI power concentration — days before a Xi-Trump summit where AI safety is on the agenda.
Key Takeaways
Liu Shengyu , DeepSeek engineer behind the V4.1 models, attacked Anthropic and OpenAI on 15 September 2026 over their calls to 'pace' AI development.
He wrote: 'I especially don't want Anthropic to master the most advanced artificial intelligence or AGI.' Liu invoked a comparison to Nazi Germany to warn against concentrating transformative AI in a single entity's hands.
The remarks precede an expected Xi Jinping–Donald Trump meeting on September 24 , where AI safety is anticipated to feature as a key topic.
DeepSeek 's open-source model strategy is positioned as a direct counter to the closed-source approaches of Anthropic and OpenAI .
The dispute intersects with US export controls on Nvidia chips and broader geopolitical competition over AI supremacy.
Liu Shengyu, a DeepSeek engineer credited with building the company's V4.1 models, has publicly attacked Anthropic and OpenAI over their calls to slow down AI development on safety grounds, framing the argument as a veiled bid by proprietary US labs to monopolise the most powerful AI systems. His remarks, posted on social media on Monday, 15 September 2026, have intensified an already charged debate in China over Western AI governance narratives.
The Core Accusation
Liu stated bluntly that he does not trust Anthropic or OpenAI to make cutting-edge AI 'open and affordable,' drawing a sharp contrast between their closed-source models and DeepSeek's open-source philosophy. 'I especially don't want Anthropic to master the most advanced artificial intelligence or AGI,' he wrote. AGI — artificial general intelligence — broadly refers to AI systems that match or surpass human-level cognitive ability.Why It Matters
The timing is geopolitically charged. The remarks came ahead of an anticipated meeting between President Xi Jinping and US President Donald Trump on September 24, where AI safety is expected to feature prominently as both nations compete for technological supremacy. The 'pacing' debate — whether frontier AI development should be deliberately slowed — has become a proxy battleground between open-source advocates and proprietary lab incumbents globally.The Nazi Germany Invocation
According to reports, Liu's post invoked a comparison to Nazi Germany to underscore the dangers of concentrating transformative technology in the hands of a single entity or ideology — a rhetorical escalation that drew significant attention across Chinese and international tech communities. The analogy reflects a broader anxiety in China that US-led AI safety discourse is structurally designed to preserve American dominance rather than address genuine existential risk.The Competitive Backdrop
DeepSeek's open-source releases have repeatedly disrupted assumptions about the cost and accessibility of frontier AI, challenging the resource-intensive, closed-model approach championed by OpenAI — backed by Microsoft — and Anthropic, which counts Amazon and Google among its major investors. The debate also intersects with ongoing US export controls on Nvidia chips, which have shaped how Chinese AI labs approach model efficiency and openness. Industry analysts have noted that 'pacing' arguments, when advanced by labs that already hold frontier positions, can function as competitive moats dressed in ethical language.What's Next
All eyes will be on the Xi-Trump summit on September 24 to gauge whether AI safety becomes a formal bilateral agenda item or remains a point of strategic friction. DeepSeek's increasingly vocal posture on AI governance signals that China's leading open-source labs are willing to engage the ideological dimension of the AI race — not just the technical one. How Anthropic and OpenAI respond, and whether Western regulators treat 'pacing' proposals as policy or competitive strategy, will define the next phase of the global AI governance debate.Point of View
Making the ideological battle as consequential as the technical one. The Nazi Germany invocation is rhetorically extreme but strategically deliberate: it reframes a US regulatory narrative as an existential power grab, a framing that will resonate inside China regardless of how it lands abroad. With the Xi-Trump summit looming, this kind of public positioning by a prominent Chinese AI engineer is unlikely to be entirely spontaneous — it shapes the negotiating atmosphere before diplomats even sit down.
NationPress
16 Sept 2026
Frequently Asked Questions
What did the DeepSeek engineer say about Anthropic and OpenAI?
Liu Shengyu , a DeepSeek engineer behind the V4.1 models, said he does not trust Anthropic or OpenAI to make advanced AI 'open and affordable' and stated he especially does not want Anthropic to control AGI. He posted the remarks on social media on 15 September 2026 .
Why did a DeepSeek engineer invoke Nazi Germany in the AI debate?
According to reports, Liu Shengyu used the comparison to warn against the dangerous concentration of transformative technology — in this case, frontier AI — in the hands of a single entity. The analogy was intended to frame Anthropic and OpenAI 's 'pacing' arguments as a cover for monopolising AGI rather than a genuine safety measure.
What is the AI 'pacing' debate and why does it matter?
The 'pacing' debate refers to calls — largely from US frontier labs like Anthropic and OpenAI — to deliberately slow AI development to allow safety research to catch up. Critics, including voices from China 's open-source AI community, argue this functions as a competitive moat that preserves the dominance of already-powerful incumbents.
How does this relate to the Xi Jinping–Trump summit?
An anticipated meeting between President Xi Jinping and US President Donald Trump on September 24, 2026 is expected to include AI safety as a key discussion point. Liu 's remarks, made days before the summit, add public pressure to the bilateral AI governance conversation.
How does DeepSeek's approach differ from Anthropic and OpenAI?
DeepSeek has embraced an open-source model strategy, releasing its models publicly, in contrast to the closed-source approaches of Anthropic and OpenAI . This distinction is central to Liu Shengyu 's argument that proprietary US labs cannot be trusted to keep advanced AI accessible and affordable.
