Zuckerberg Tells FT US Should Not Ban Chinese AI Models

Mark Zuckerberg used a pair of rare interviews, with the Financial Times and the New York Times, to argue that the US should not try to gain an AI edge by blocking Chinese models from American use. He called that approach "is not an effective solution" and pointed at OpenAI and Anthropic, saying rivals are trying to sway government regulation in ways that concentrate power in a handful of closed labs.

The frame he keeps returning to is that "The broad trend of the entire industry's development has been towards greater openness, putting more technological power in the hands of more people, not fewer." That is a convenient position for the company shipping one of the most widely used open-weight families of models, but it is also a live policy fight. Reporting around the interviews notes that OpenAI and Anthropic support up to 30 days of federal access before some frontier model releases, a proposed process that would apply to Meta and xAI as well.

The strategic subtext is worth naming. If Chinese open-weight releases like Moonshot's Kimi K3, put out on July 16, keep closing the capability gap, the US case for restricting them starts to look less like a competitive moat and more like a wall around a shrinking garden. Zuckerberg's bet, as summarized in secondary coverage of the FT interview, is that "the democratization of technological tools ultimately tends to win out," and Meta would rather compete on distribution and infrastructure than on gate-keeping through Washington.

The honest caveat is that the interview coverage circulating outside the FT paywall is thin on exact wording, and it does not spell out what enforcement Zuckerberg would actually accept in place of a ban, or how he squares this position with Meta's broader posture toward China. Take the specifics as reported, not settled.

The interesting question over the next few months is whether the open-weight camp can turn this from a CEO talking point into a coalition. A letter titled "Open Weights and American AI Leadership" was signed by 25 organizations including Nvidia, Microsoft and Meta, notably without OpenAI or Anthropic on it. If that group can offer a credible alternative to the closed-labs review regime, the US AI policy debate stops being a two-horse race.