Zuckerberg, Sanders Clash Over U.S. AI Regulation

Mark Zuckerberg and Senator Bernie Sanders took opposing positions on U.S. artificial intelligence regulation on Thursday, setting up a debate in Washington over whether development should continue while rules are written or be paused until stricter federal controls are in place.

Zuckerberg, chief executive of Meta, warned that a national regulator modeled on the Financial Industry Regulatory Authority would slow U.S. releases of advanced AI models and allow foreign competitors to gain an edge. He raised the point in a private call with President Trump in August and has argued that technologies approaching superintelligence should be broadly available rather than concentrated in a few labs.

Sanders, joined by Representative Greg Casar, introduced the Ban Artificial Superintelligence Act. The bill would prohibit AI systems that match or exceed human cognitive performance, freeze advanced AI development until a federal regulator issues rules, and impose penalties including up to 20 years in prison for violations. Sanders wrote in a statement: “The future of humanity cannot be left in the hands of a handful of Big Tech oligarchs.”

One regulatory option under discussion is a FINRA-style federal body to test frontier AI models for safety before release. The idea was promoted earlier this year by Google DeepMind chief Demis Hassabis. Supporters say pre-release testing could identify risks in powerful models.

Zuckerberg and some advisers prefer lighter-touch approaches such as industry self-regulation or a voluntary framework, arguing heavy pre-release controls could slow innovation. Adviser David Sacks has dismissed some AI safety concerns as exaggerated storytelling.

Lawmakers, company executives and researchers are negotiating whether any referee should require a temporary halt to model development while rules are established. Supporters of legal limits contend the possible harms of advanced AI justify a freeze; opponents argue delays could cede technological advantage to other countries.

Proposals on the table range from voluntary industry groups and independent audits to mandatory testing regimes and criminal penalties. If enacted as written, the Sanders bill would ban systems once they reach human-level cognitive capabilities.