Live Updates > Live Update Details
2026-09-17 13:16:09
[Zuckerberg refutes “slowdown theory”: AI safety does not require external pressure, labs that do not focus on alignment will fall behind] (1) Meta CEO Mark Zuckerberg said on Tuesday that AI companies are inherently motivated to focus on safety when deciding on the timeline for building models, and do not need external pressure to maintain this timeline. (2) He posted on the social platform X that every lab has the responsibility and motivation to move at the pace required to safely train its models, and the ability to take action on its own to ensure that this is achieved. (3) Zuckerberg also asserted that people do not want to use agents that are inconsistent with them and do not act as they ask, which puts pressure on AI creators to make their models more in line with human values. (4) He added that there is a lot of debate about slowing down progress on capabilities before “alignment” catches up; his point is that trust and alignment are rapidly becoming the most important capabilities that differentiate agents and models, and any lab that does not focus on alignment will fall behind. (5) Zuckerberg’s post came after a wider debate surrounding AI development over the past week. Anthropic CEO Dario Amodei said over the weekend that the development of this technology must be slowed down to allow room for the implementation of necessary risk prevention and mitigation measures; tech peers Sam Altman and Elon Musk both support Amodei. (6) In a blog post, Amodei wrote: “We must slow down the pace of improving the capabilities of AI models. Progress will still seem fast, and we must use the time we gain wisely.” (7) The Anthropic CEO suggested that AI companies should quickly take steps to “adjust” the pace of model development and warned that if the technology is not put on the brakes, it could have serious consequences.
