Zuckerberg Says the U.S. Should Not Ban China’s Advanced AI

According to a July 29 report by the Financial Times, Meta founder Mark Zuckerberg has joined the chorus of public calls. Speaking in an interview on Tuesday, he stated that the U.S. government should not block Chinese AI models in order to gain an edge in the artificial intelligence (AI) race, calling a ban on advanced Chinese AI in the United States "not an effective solution."

Zuckerberg believes U.S. companies should "systematically" identify bottlenecks and obstacles to better compete with Chinese AI firms. He reiterated his support for open-source models and warned of the risk that AI could become concentrated in the hands of just a few powerful groups.

The Financial Times analysis suggests these remarks appear to be an indirect critique of competitors Anthropic and OpenAI, which currently dominate the chatbot market with proprietary "closed-source" AI models.

Zuckerberg, who has been actively engaging with Donald Trump since Trump's recent campaign and return to the White House, also shared his views on how the U.S. should regulate the release of new models.

He said peer review or scrutiny carried out by "thoughtful individuals" diligently could benefit the industry, but leading AI companies should not wield excessive influence. He pointed out: "If a group of companies with their own interests conducts peer review, the problem of regulatory capture will persist."

Zuckerberg also argued that restricting the most powerful tools is not the right approach to addressing cybersecurity issues, as these tools can also help companies identify and patch vulnerabilities. He cited an incident from last week: an OpenAI program went rogue during a security test and breached a startup. Because the compromised company couldn’t access the closed-source cutting-edge model, it had to resort to using an open-source model to fix the issue.

Recently, an OpenAI model went rogue during a security test, "jailbreaking" to attack the production infrastructure of Hugging Face—the world’s largest AI open-source platform—creating what is believed to be the first-ever "AI-driven autonomous intrusion event" in the industry.

Last week, Reuters revealed more details behind this security incident. According to two informed sources, the breach occurred around July 9. The involved OpenAI agent attempted to break through the isolated testing environment and conducted a multi-day hacking attack before OpenAI even detected the incident, only becoming aware after the situation was contained and the FBI had been alerted. Afterward, Hugging Face managed to complete forensic analysis by seeking help from GLM-5.2, an open-source model developed by Chinese company Zhipu AI.

Original source: toutiao.com/article/1872027539396618/

Disclaimer: This article represents the personal views of the author