In a post on X, Zuckerberg said users will not want to use AI agents that are not aligned with their interests or that fail to properly carry out their instructions.
According to the Meta founder, this creates a natural incentive for companies to invest in safety and continuously improve how their models respond to users.
Zuckerberg also said Meta had delayed the launch of its Muse AI system for several months in order to strengthen safety measures before releasing it.
He also defended the use of external experts to evaluate artificial intelligence systems and argued that companies should continue developing the technology without stopping its progress.
His position comes as other major technology companies discuss ways to strengthen oversight of the most advanced AI models.
OpenAI, Anthropic and Google DeepMind are working on the creation of a joint body to oversee risks associated with artificial intelligence and establish safety standards for the industry.
The proposal could require the most advanced AI models to undergo safety assessments before being widely released, particularly to identify potential risks to users.
The idea has gained attention as several industry leaders argue that AI development should move at a pace that allows safety measures to keep up with the technology's growing capabilities.
Zuckerberg, however, rejects the idea of a coordinated pause or slowdown among companies. He argues that each AI laboratory should take responsibility for the safety of its own systems without stopping the development of artificial intelligence.
The debate among the world's leading technology companies is now focused not only on how quickly AI should advance, but also on how its risks should be managed.