Meta has unveiled its new agentic AI model, Muse Glimmer. Business Insider reported on Aug. 11 that Mark Zuckerberg (마크 저커버그) set out a vision of “AI for everyone.” The model is released as open weights, allowing users to download it to their computers and modify it for free. Zuckerberg, describing the future of AI, argued there is no need to fear AI too much.

The industry debate over whether to keep AI models open or run them as closed systems is also continuing. Supporters of open models see it as beneficial for the broader industry for developers to be able to adjust a model. Supporters of closed models argue AI technology is too powerful to allow broad access.

Both lines of argument have limits. Companies without top-tier models can more easily support opening models without losing much competitive advantage. On the other hand, even leading AI companies are struggling to control core models, raising the question of whether they can argue they alone should manage models they cannot control themselves.

A bigger obstacle to Zuckerberg’s vision is public acceptance. Users are reacting strongly not only to AI but also to the data centres that support it.

Zuckerberg has presented a vision for personal superintelligent AI that pursues goals, manages information and even takes actions on a user’s behalf. CXToday reported the concept includes customer-side AI agents filtering brand messages, evaluating offers and handling repetitive interactions in place of customers.

In a post on X, Zuckerberg said superintelligence should be provided to individuals rather than concentrated in governments, companies or institutions. He said everyone should be able to access superintelligence, and personal AI should understand a user’s goals, interests and context to help with decision-making, creation, learning, information management and carrying out future actions.

If the vision becomes reality, the starting point of the customer journey could be personal AI rather than brands. The structure would have personal agents first review product information, prices, policies, availability and services and decide which interactions require a person’s attention.

Meta holds a large amount of personal-related data, leaving room to develop AI into a more personalised assistant. Ultimately, the success or failure of Meta’s strategy depends not only on whether it opens its models but on whether users can first become accustomed to using AI. The move shows that opening AI models, infrastructure acceptance and user expansion strategies are tied together.

Technical performance alone is not enough, and for Meta as a mass-market services company, changing user habits has emerged as the central task of its AI strategy.