Meta Platforms CEO Mark Zuckerberg stated that competition and accountability issues give AI development companies sufficient grounds to act independently regarding safety. His statement effectively contradicts calls from leaders of leading AI companies for a coordinated slowdown in artificial intelligence development.
These comments highlight a division among industry leaders and US authorities regarding how to respond to warnings that the technology is developing faster than companies can control it, while some industry insiders warn of an existential threat to humanity.
In a post on X on Tuesday, Zuckerberg noted that each lab bears responsibility and has the incentive to move at the required speed for safely training its models, and also possesses the ability to take its own measures to ensure this.
His remarks came three days after Anthropic CEO Dario Amodei called on AI companies to slow down the pace of model capability improvements, following warnings about existential risk from several researchers. Competitors, such as OpenAI head Sam Altman and Elon Musk of xAI, supported Amodei's call within hours.
However, US President Donald Trump downplayed concerns about potential AI harm on Monday, stating that protective mechanisms already exist for this technology in the US, and that China would benefit from doubts about AI development. The view that regulation would hinder American companies' ability to compete with their Chinese counterparts is shared by tech executives, including Nvidia CEO Jensen Huang.
Zuckerberg added: 'Trust and alignment are quickly becoming the most important capabilities that will distinguish agents and models. Any lab that does not focus on alignment will fall behind.'
He also reported that labs have a strong incentive to prevent models from causing harm, as otherwise they will face 'significant' liability. He cited Meta's decision to delay the release of its Muse AI agent earlier this year to enhance safety.
Zuckerberg explained: 'We did not call on everyone else to do it before us. We simply did it as part of our daily work because it was obviously right for people and for us.'
While Amodei advocated for a cross-industry slowdown of recursive self-improvement—the ability of AI to improve itself—Zuckerberg stated that Meta has already directed 'a significant majority' of its computing power toward creating products that meet immediate user needs, rather than developing self-improving AI systems.
He mentioned that Meta's Superintelligence Labs division, which is the company's AI department, already uses independent evaluators in several areas, calling it 'industry best practice,' and added that 'other labs can just do the same.'
Zuckerberg's attempt to portray Meta as a responsible player in AI safety came several weeks after the company agreed to pay up to $18 billion to settle lawsuits from US states alleging that Meta designed Facebook and Instagram in a way that caused child addiction, without admitting guilt.
Heightening uncertainty about any coordinated approach from AI firms, US Federal Trade Commission Chairman Andrew Ferguson stated on Tuesday that everyone should be 'deeply skeptical' of AI companies seeking exemption from antitrust legislation while lobbying for new regulation. Ferguson did not name Anthropic, but last week Amodei requested antitrust regulatory approval to allow competitors to coordinate a slower pace of AI development for safety reasons.