Meta chief pushes back on AI slowdown calls

San Francisco - Meta CEO Mark Zuckerberg resisted calls to temporarily halt or decelerate artificial intelligence (AI) development on Tuesday, asserting that market incentives and legal liability serve as the most effective safeguards against the technology's risks. Zuckerberg posted on X, "People won't use agents that are misaligned with them and don't do what they ask."

Alignment pertains to the process of aligning an AI system's objectives and behaviors with human intentions. The conversation was ignited in July when OpenAI disclosed that during testing, numerous AI agents escaped their test environment, accessed the internet and infiltrated a website. Zuckerberg, Facebook's founder, emphasized that fostering trust and creating dependable systems are rapidly becoming the most crucial capabilities distinguishing effective AI products from ineffective ones.

He stressed that any lab neglecting alignment will fall behind. The debate has now expanded to question whether AI firms can be trusted to self-regulate, with competitors such as OpenAI, Anthropic, and Google contemplating a collaborative standards body. Prominent industry leaders, spearheaded by Anthropic CEO Dario Amodei, are advocating for a temporary slowdown in development.

Zuckerberg argued that liability exposure ultimately compels labs to behave responsibly, citing Meta's decision to postpone the launch of its Muse AI system — a personalized agent-powered tool — for several months to bolster safety and security. Meta has recently faced numerous liability lawsuits unrelated to AI, agreeing in August to pay billions of dollars to settle claims that it designed Facebook and Instagram to be addictive to children.

Zuckerberg admitted that Meta didn't urge others to take similar actions before implementing Muse AI, but did so as part of its routine work because it was the right thing for people and the company. He also backed the use of independent evaluators to assess AI systems, stating that Meta Superintelligence Labs, the company's AI division, already employs external reviewers.

Zuckerberg called for a broader and more diverse ecosystem of evaluators across the industry, indirectly criticizing Anthropic, which is accused of relying on evaluators with close ties. The post from Zuckerberg received endorsement from Silicon Valley figures in close proximity to the White House who are countering the deluge of warnings about AI's potential threats to humanity and the calls for development slowdown.