Meta delays Muse AI models as Zuckerberg cites safety concerns

17:00

By: Azzat Manal

Meta delays Muse AI models as Zuckerberg cites safety concerns

Meta has postponed the release of its new Muse artificial intelligence models by several months, with CEO Mark Zuckerberg saying the additional time will be used to address safety and security concerns surrounding advanced AI systems.

In a post on X on Wednesday, Zuckerberg said AI companies and research laboratories have a responsibility to determine the pace of development and deployment of their models in a way that allows sufficient time for testing and risk assessment before wider public release.

The Meta chief said the decision to delay Muse was made independently by the company rather than in response to pressure from competitors or external organizations. He stressed that safety measures should be incorporated into the development process from the beginning rather than treated as a final step before a model reaches users.

Zuckerberg also highlighted the role of independent evaluations in assessing advanced AI systems. External testing, he argued, can help identify potential risks that may not be detected through a company's internal evaluation procedures.

He called for broader and more diverse independent assessment efforts, involving a wider range of experts and organizations. Such evaluations could provide additional scrutiny of increasingly capable AI models before they are deployed on a large scale.

The decision comes as major technology companies continue to compete in the rapidly evolving AI market, with increasingly sophisticated systems being developed for a growing range of applications. Alongside advances in performance, concerns have intensified over potential misuse, cybersecurity vulnerabilities and the difficulty of predicting certain outputs from advanced models.

Meta's decision to postpone Muse highlights the growing emphasis on managing AI risks within technology companies themselves, while also strengthening the role of external oversight and independent testing.

The move also reflects a broader challenge facing the industry: balancing the pressure to release increasingly capable AI systems with the need to give developers enough time to evaluate their capabilities, limitations and potential risks before they become widely available.