Meta’s Zuckerberg says AI labs have enough incentive to build safely

On Tuesday, Meta platform CEO Mark Zuckerberg stated that AI companies have sufficient incentive to independently focus on safety through competition and liability. This stance contrasts with calls from leaders of top AI firms for a collective slowing of AI development. The remarks highlight a division among industry executives and U.S. authorities regarding how to respond to warnings that AI technology is advancing too rapidly and could pose existential risks to humanity.

Zuckerberg asserted that every lab has a responsibility and incentive to train models safely at the necessary pace. His comments follow three days after Anthropic CEO Dario Amodei urged AI companies to slow down model improvements due to existential risks. Rival OpenAI CEO Sam Altman and xAI founder Elon Musk both supported Amodei's call shortly after.

President Trump, however, downplayed AI safety concerns, stating that the U.S. already has safeguards in place and that China would benefit from skepticism towards AI development. Tech executives, including Nvidia CEO Jensen Huang, echoed this view, saying that trust and alignment are crucial for differentiating AI agents and models.

Meta CEO Zuckerberg emphasized that labs have a strong incentive to prevent their models from causing harm, as failure to do so could result in significant liability. He cited Meta's decision to delay the release of its Muse AI agent earlier this year to improve security.

While Amodei called for an industry-wide reduction in recursive self-improvement, Zuckerberg noted that Meta has already allocated the majority of its computing power to creating user-focused products rather than pursuing self-improving AI systems. Meta's AI division, Superintelligence Labs, also employs independent evaluators, describing it as "industry best practice."

Zuckerberg's stance aims to position Meta as a responsible actor in AI safety, following the company's recent $18 billion settlement with U.S. states over allegations that Facebook and Instagram were designed to addict children. This decision did not involve admitting fault.

Adding to the uncertainty over any coordinated approach by AI firms, FTC Chairman Andrew Ferguson warned that everyone should be "deeply suspicious" of companies seeking antitrust exemptions while advocating for new regulations. Ferguson did not specifically mention Anthropic but noted that Amodei had requested antitrust permission to enable rivals to coordinate a slower pace of AI development due to safety concerns.