US AI LEADERS TURN TO CHINESE OPEN-WEIGHT MODELS, CHALLENGING CLOSED-SOURCE SAFETY CLAIMS

AI pioneer Andrew Ng, a former head of Google Brain and Baidu’s chief scientist, publicly stated that Chinese open-weight models appear safer than closed-source alternatives, a claim that directly challenges the long‑held position of U.S. firms like Anthropic that argue open-source models pose a societal threat. His remarks come in a broader debate over AI safety and control in the United States.

According to the South China Morning Post, Ng said, “From what I’m seeing, I think open-weight models seem safer to me than closed-weight models.” The comment follows Anthropic’s assertion that open-source models are a danger to society, a narrative that has dominated U.S. AI policy discussions. Ng’s observation underscores the growing influence of Chinese AI technology, which has been described in other reports as driving a shift in global AI dynamics.

The United States has historically promoted closed-source AI as a more secure approach, while Chinese firms increasingly release large language models that are open and weighable, raising questions about the best path to responsible AI development. This debate reflects broader tensions highlighted in the Guardian’s coverage of China’s tech advances affecting both Silicon Valley and U.S. policy circles.

• Andrew Ng endorses Chinese open-weight AI models as safer than closed-source ones.
• The stance counters Anthropic’s claim that open-source models threaten society.
• The remarks align with growing media scrutiny of China’s AI progress and its impact on global AI policy.

Read the original source for full detail available at News - South China Morning Post src