Meta Rolls Out Muse Wonder AI With Persistent Memory To Subscribers And API Users

Meta has launched two new artificial intelligence models, Muse Wonder and Muse Wonder Full, marking a significant milestone in its ongoing AI development efforts. According to Engadget, the company states that these models represent a major step toward personal superintelligence after more than a year of rebuilding its AI infrastructure. The rollout targets paid subscribers and API users, with specific pricing and tier requirements detailed in the release.

The risk

The introduction of active memory and active focus capabilities changes how these AI agents operate over extended periods. Active memory allows the system to retain relevant details across multiple conversations, while active focus enables the agent to persist through difficult tasks. For organizations, this means the AI retains context that could include sensitive operational data, legal strategies, or financial information. The ability to carry data across sessions increases the surface area for potential data leakage or unintended information retention if not properly managed.

Who is exposed

Muse Wonder is currently available through Meta’s API and within the Muse Code platform, according to Engadget. Muse Wonder Full is restricted to users on the $50 per month power usage tier or the $200 per month maximum usage tier of Muse Code. The broader Muse personal agent will receive these capabilities starting with paid subscribers. Developers and businesses integrating these models via API are directly exposed to the new data handling behaviors, as are individual subscribers using the higher-tier coding tools.

Cost and compliance

Meta reports that the Muse Wonder model costs $4 per million input tokens and $20 per million output tokens when accessed via API. The Muse Wonder Full model requires a subscription to the $50 or $200 monthly tiers of Muse Code. CEO Mark Zuckerberg told the Financial Times that reaching this development stage required substantial compute resources and rigorous scientific work. Meta expects to spend up to $145 billion this year on AI initiatives, indicating a massive financial commitment to maintaining competitive positioning in the frontier AI market.

What should you do?

Business leaders should review how these new memory and focus features interact with their existing data governance policies. Organizations using the API must ensure that the persistent memory function does not inadvertently store or retrieve sensitive proprietary information across different user sessions. It is critical to monitor the specific data retention behaviors of the active memory feature to maintain compliance with internal security standards and external regulatory requirements regarding data privacy and handling.

Background: AI agents

AI agents are software systems that perceive their environment, make decisions, and execute actions to achieve specific goals. Unlike standard models, they maintain internal state and interact with external tools, creating unique security risks around tool execution, memory retention, and autonomous decision loops that require specialized controls.

Read the full guide: AI Agents Explained: Answers to Security and Architecture Questions

What to do and how to stay safe: Meta

• Audit API integrations to ensure active memory features do not store sensitive data beyond necessary session limits.
• Define clear data retention policies for AI agents that persist context across multiple interactions.
• Monitor usage costs and token consumption closely, as persistent memory may increase output token usage.
• Verify that subscription tiers align with the organization's security requirements and data handling capabilities.

Step-by-step guide: AI Governance Explained: Building Rules for Autonomous Systems

General security guidance from the Vector Tribune newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

What are active memory and active focus in Meta's new models?

Active memory allows the AI to retain relevant details across conversations, while active focus helps it persist through difficult tasks toward a specific goal.

How much does the Muse Wonder API cost?

The Muse Wonder model is priced at $4 per million input tokens and $20 per million output tokens according to Meta's pricing structure.

Who can access the Muse Wonder Full model?

Muse Wonder Full is available to subscribers of the $50/month power usage or $200/month maximum usage tiers of Muse Code.