Mark Zuckerberg, the boss of Meta, on Tuesday rejected growing calls to suspend or slow down the development of artificial intelligence, believing that market mechanisms and the risk of legal proceedings constituted the best safeguards against this technology.
“People won’t want to use agents who aren’t aligned with their interests and don’t do what they ask. Labs therefore have a strong natural incentive to make their models more aligned,” wrote Mark Zuckerberg on X.
“Alignment” refers to efforts to make models behave the way humans want them to.
“Any lab that doesn’t focus on alignment will fall behind,” he added.
Concern about AI has grown in recent weeks, due to several security incidents and the new propensity of artificial intelligence models to improve on their own, without human intervention.
The debate today also concerns the sector’s capacity to self-regulate. Players like OpenAI, Anthropic and Google are exploring the creation of a joint body responsible for developing standards.
Several influential figures in the sector, led by the general director of Anthropic, Dario Amodei, are calling for a slowdown in the development of AI.
Mr. Zuckerberg, however, defends that the threat of legal action already forces laboratories to act responsibly.
He cites as an example Meta’s decision to postpone the launch of its Muse AI system, equipped with personalized agents, by several months in order to strengthen its security and safety systems.
Mark Zuckerberg also spoke in favor of independent evaluations of AI systems, pointing out that Meta Superintelligence Labs, the group’s AI division, already uses outside experts.
He called for a “broader, more diverse ecosystem of reviewers” across the industry, a thinly veiled criticism of Anthropic, which some have accused of favoring reviewers with whom the company has close ties.
Meta, which is investing up to $145 billion this year in its AI infrastructures, is fighting against any regulation likely to slow it down: no rule should delay the release of American models, “even if only for a month,” its founder had already pleaded at the beginning of August.
Mark Zuckerberg’s publication was welcomed by several Silicon Valley figures close to the White House, who contest the multiplication of warnings about the potential dangers of AI and calls for a slowdown in its development.
