Zuckerberg Says Competition and Liability Drive AI Safety, Splitting Industry
Mark Zuckerberg argues competition and liability give AI labs enough incentive to build safely, as Meta reveals it delayed its Muse agent over security.
Meta Platforms CEO Mark Zuckerberg has argued that competition and legal liability give artificial intelligence companies sufficient reason to act on safety individually, a position that appears to diverge from appeals by leaders of rival AI firms for a coordinated slowdown in development.
In a post on X on Tuesday, Zuckerberg said every laboratory has both the responsibility and the incentive to move at whatever pace is needed to train its models safely, and the ability to take its own steps to ensure that happens. He added that trust and alignment are becoming the most important capabilities distinguishing AI agents and models, warning that any lab neglecting alignment will fall behind.
His comments came three days after Anthropic CEO Dario Amodei urged AI companies to slow the pace at which they improve model capabilities, following warnings from several researchers about existential risks. OpenAI chief Sam Altman and xAI's Elon Musk publicly backed Amodei's call within hours.
Zuckerberg said labs have a strong incentive to prevent their models from causing harm because they would otherwise face significant liability. He cited Meta's decision earlier this year to delay the release of its Muse AI agent in order to strengthen security, saying the company did not wait for others to act first but treated it as part of its routine work because it was the right thing for users and for the company.
On recursive self-improvement — AI's ability to improve itself, the focus of Amodei's appeal — Zuckerberg said Meta had already committed the significant majority of its compute to products serving users' immediate needs rather than to self-improving systems. He also said Meta Superintelligence Labs, the company's AI division, already engages independent evaluators in several areas, describing the practice as an industry best practice that other labs can adopt.
The debate over how to handle AI risks has exposed differences among industry executives and U.S. authorities. President Donald Trump on Monday played down concerns about AI's potential for harm, saying the United States already had guardrails in place and that China would benefit if doubt were cast on AI development. That view — that regulation could hurt American firms' ability to compete with Chinese counterparts — is shared by tech executives including Nvidia CEO Jensen Huang.
Adding to the uncertainty over any coordinated industry approach, Federal Trade Commission Chairman Andrew Ferguson said on Tuesday that everyone should be deeply suspicious of AI companies seeking antitrust exemptions while lobbying for new regulations. Ferguson did not name Anthropic, but Amodei last week requested antitrust permission to let rivals coordinate a slower pace of AI development on safety grounds.
Zuckerberg's effort to cast Meta as a responsible actor on AI safety comes weeks after the company agreed to pay up to $18 billion to settle U.S. state lawsuits alleging it designed Facebook and Instagram to addict children, without admitting wrongdoing.