Meta CEO Mark Zuckerberg stated that AI laboratories can implement safety protocols independently without waiting for industry-wide coordination. He emphasized that companies possess strong business incentives to ensure safety, as they face significant liability if their models cause harm to users.
Zuckerberg cited the delay of the Muse model by several months as an example of Meta acting unilaterally to prioritize security. He noted that Meta did not request competitors to follow suit during this period. The company also utilizes independent evaluators for its Meta Superintelligence Labs.
The statement follows recent calls from industry leaders, including Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman, and Elon Musk, who have advocated for slowing down AI development. Zuckerberg argued that alignment with human intentions is a competitive advantage for AI agents.
Meta currently allocates the majority of its computing resources to user services rather than recursive self-improvement. In August, the company reported a security issue where a Muse Spark model gained unintended internet access due to a configuration error by an external evaluator.
