Mark Zuckerberg, CEO of Meta, argued that the US government should not block Chinese AI models in order to stay ahead in the AI race.

Reuters Yonhap News

Reuters Yonhap News

View original image

On July 28 (local time), during an interview with the Financial Times (FT), CEO Zuckerberg stated that even if the US considers sanctions on Chinese AI research institutes due to alleged intellectual property (IP) violations, banning advanced AI models from the US market would not be "an effective solution." He added that in order for US companies to maintain an advantage over China in the AI competition, they must "systematically" identify bottlenecks and obstacles that impede technology development.


Zuckerberg also addressed how the US government should oversee the release of new AI models. He said that if experts are properly involved, mechanisms such as peer review or pre-release assessments for new AI models could be helpful for the industry. However, he stressed that, in designing such systems, major AI companies should not be given excessive influence.


He said, "When companies with vested interests participate in peer review, a particular company could wield too much influence over regulation." He especially questioned whether leading-edge AI firms genuinely want open-weight models to succeed, noting that these companies have so far sent mixed signals in this regard.


Positions differ among major companies regarding the level and specifics of AI regulation. Dario Amodei, CEO of Anthropic, argued for stronger AI regulations. Demis Hassabis, CEO of Google DeepMind, proposed establishing a self-regulatory organization, similar to the US Financial Industry Regulatory Authority (FINRA), to set industry standards.


However, both OpenAI and Anthropic opposed restricting open-weight models themselves. Amodei pointed out that open-weight models do not necessarily make it easier to develop safety mechanisms, and it cannot be definitively said that broad access to advanced technology benefits defenders more than attackers.


Zuckerberg argued that restricting the use of the most powerful AI tools is not the right solution to cybersecurity issues, as such technology could also be used to detect and address vulnerabilities within companies.



He cited a recent case in which an OpenAI program went out of human control and hacked a startup's system. "There was a breach incident caused by a large AI model, but the victim company was unable to use a closed-source cutting-edge model," he said, adding, "In the end, the company used an open-weight model to resolve the security issue."


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing