DeepSeek Engineer Defends Open-Source AGI Push Despite Safety Risks

A Chinese AI laboratory's public commitment to releasing human-level artificial intelligence as open-source code places it in direct tension with both US export controls and China's own emerging...

A DeepSeek engineer said Tuesday that the Chinese AI laboratory will continue developing and releasing open-source artificial intelligence aimed at achieving artificial general intelligence (AGI), even if that approach increases safety risks, according to the South China Morning Post.

The engineer made the statement during an AI panel at the World Internet Conference (WIC) held in Wuzhen, China, on Tuesday, November 4, 2026, according to the South China Morning Post. The WIC is an annual state-affiliated technology forum that draws participants from major Chinese and international technology companies.

AGI refers to AI systems capable of performing any cognitive task a human can perform, a threshold no system has yet demonstrably reached. The term is distinct from narrow AI, which performs specific tasks. The timeline for reaching AGI is contested among researchers, with estimates ranging from years to decades, according to published assessments from organizations including the Machine Intelligence Research Institute and OpenAI.

DeepSeek drew international attention in early 2025 when it released its R1 reasoning model as open-source software, allowing any developer to download, modify, and deploy it without licensing fees. The release prompted significant discussion in the United States about whether export control policies were sufficient to limit the spread of advanced AI capabilities developed in China, according to reporting by multiple US technology publications at the time.

Open-source AI release means the underlying model weights, the numerical parameters that define a model's behavior, are made publicly available. Once released, those weights cannot be retracted, which is the core of the safety concern raised by critics: a model with dangerous capabilities, once published, is effectively uncontrollable.

The engineer's remarks at WIC defied scrutiny from two directions simultaneously. US officials and legislators have raised concerns that open-source AI releases by Chinese firms could allow adversaries or non-state actors to access frontier capabilities. Separately, Chinese regulators have been developing their own AI governance rules under the Interim Measures for the Management of Generative Artificial Intelligence Services, which took effect in August 2023, creating a domestic review framework whose application to open-source AGI-class models remains untested.

DeepSeek has not published a formal safety policy equivalent to those released by US-based frontier AI labs such as Anthropic or OpenAI, whose published documents outline risk thresholds and evaluation procedures before major model releases. It is not publicly known what internal safety evaluations, if any, DeepSeek conducts before releasing model weights.

The company's operational structure also remains partially opaque. DeepSeek is affiliated with the Chinese quantitative hedge fund High-Flyer Capital Management. The relationship between the fund's investment objectives and the laboratory's research agenda has not been disclosed in public filings or statements, according to available corporate records.

The WIC forum itself has historically been used by Chinese officials to articulate the government's position on internet governance and technology policy. The choice of that venue for the engineer's remarks gives the statement a degree of institutional visibility beyond a typical corporate announcement, though the engineer spoke as an individual panelist rather than issuing a formal company policy document.

The specific capabilities of DeepSeek's current and planned models, including whether any internal version approaches AGI-level performance on standard benchmarks, were not detailed in the engineer's public remarks as reported by the South China Morning Post. The benchmarks used to define AGI-level performance, including the Abstraction and Reasoning Corpus and others maintained by independent researchers, remain subject to debate within the AI research community regarding their adequacy as tests of genuine general intelligence.

What would clarify the policy implications of DeepSeek's position is a formal response from either China's Cyberspace Administration or the US Commerce Department's Bureau of Industry and Security, neither of which had issued a public statement on the WIC remarks as of the time of publication.