How fast should artificial intelligence move?
That question is becoming harder for the technology industry to avoid as AI systems grow more capable, and the people building them are no longer speaking with one voice.
Anthropic CEO Dario Amodei recently called for a slower pace of development because he believes the industry needs more time to deal with the risks created by increasingly powerful AI. Sam Altman of OpenAI and Elon Musk of xAI have backed his position.
Mark Zuckerberg is pushing in the other direction.
The Meta CEO argues that companies do not need an industry-wide agreement to slow down. In his view, each AI lab already has strong reasons to make sure its systems do not cause serious harm.
Every lab has the responsibility and incentive to move at the pace required to train its models safely.
Zuckerberg’s case rests partly on what companies stand to lose. A dangerous AI system could expose its developer to serious legal liability, creating a financial reason to identify problems and strengthen safeguards.
Meta’s Muse AI agent gives him a real example to point to. The company delayed its release while it worked on security, showing that Zuckerberg’s position does not mean every AI product has to move forward at maximum speed.
We didn’t call for everyone else to do this before we would. We just did it as part of our day-to-day work.
Meta also uses independent evaluators to examine parts of its AI systems. Zuckerberg described that outside scrutiny as an industry best practice rather than something that requires companies to collectively slow their development.
For now, the argument has no easy resolution. One side wants more time between major advances, while the other believes companies can keep pushing the technology forward and build the necessary safeguards alongside it.