Manus has said it will soon resume operating as an independent company after separating from Meta, ending a corporate arrangement that began with Meta’s acquisition of the AI-agent company on December 29, 2025. In a notice to users, Manus said the transition requires a temporary data deletion, backup, and restoration process for certain users in specific jurisdictions. It said the process is necessary to comply with regulatory requirements and explicitly stated that it is not the result of a data breach or security incident.
The change is a corporate story, but it is also a revealing example of the complications surrounding cross-border AI companies. AI agents are not merely software features. They can store user prompts, documents, task histories, outputs, and workflow context. A shift in ownership or operating structure can therefore trigger questions about where data is held, who controls it, which laws apply, and how user access is maintained. Manus says its data is stored in the United States and Singapore, while the transition back to independence will affect only some accounts.
The company’s notice gives a rare operational view of how a regulatory separation can reach end users. Data generated by affected users on or after December 29, 2025 will be deleted during August 23 and 24, Singapore time. Affected users can use backup tools before the deadline and restore their data beginning August 25. Unaffected users, Manus said, will continue using the product normally. The episode adds a new dimension to the AI agent market, where product capability is increasingly inseparable from trust, data stewardship, and corporate jurisdiction.
The news arrives amid a wider race to turn large-language models into systems that can carry out real tasks. EastFrontier has followed this shift through its coverage of AI-agent platforms racing to localize in China. Manus’ separation shows that, for agent companies operating across borders, organizational structure can become as consequential as model quality.
The user notice sets out a narrow but disruptive transition
Manus’ message is specific about the schedule. Affected users can back up data until 7:59 a.m. Singapore time on August 23. The deletion period runs from 8:00 a.m. on August 23 through August 24. Restoration opens at 8:00 a.m. Singapore time on August 25. During the two-day transition, affected users will not be able to access their accounts after the backup window closes.
The company says users can create multiple backups, an important detail for people whose task histories continue to change during the notice period. If a user generates new data after making an initial backup, Manus advises making another backup so the latest material is preserved. The product also says it will notify affected users by email and in-app messaging. Users who registered with Apple ID or Facebook are told to check in-app notifications because Manus may not have an email address for them.
Those operational details make the transition more than a headline about ownership. They show that the separation affects the persistence of user-created material. For an AI agent, task data can be more sensitive than an ordinary chat log. It may include research outputs, files, automation records, code, business plans, or personal workflows. A backup-and-restoration process is therefore a significant product event even if it is temporary.
Manus has tried to limit the impact. It says unaffected accounts do not need to take action and can continue using the service. It also says affected users will not be charged during the backup period and will receive welcome-back bonuses after restoration. These measures do not eliminate inconvenience, but they acknowledge that data continuity is central to the value of an agent platform.
The separation highlights the data challenge for cross-border AI
The company attributes the process to its transition to independent operations and compliance requirements in certain jurisdictions. It does not identify a government directive, specify the legal rule involved, or explain why only data generated after the Meta acquisition date is affected. Those limits matter. It would be unsupported to characterize the event as a decision by a named regulator or a judgment on the product’s security.
Yet the general lesson is clear. AI services operate across data rules that are often national, while their users and infrastructure are global. A company can host data in one country, have corporate links in another, serve users in many jurisdictions, and rely on models or cloud services from still more places. When ownership changes, these overlapping obligations can become visible in ways that ordinary consumers experience directly.
This is particularly acute for AI agents because their usefulness comes from accumulating context. A basic search product can function with a short-lived query. An agent becomes more valuable when it remembers tasks, connects tools, manages files, and learns user preferences. That persistence also creates greater responsibility. Companies must explain what happens to stored material when terms, ownership, infrastructure, or regulatory exposure changes.
Manus’ statement that its data is stored in the United States and Singapore provides one part of the picture. Its decision to give users a backup period provides another. Together, they show an attempt to preserve user agency during a structural transition. But they also underline the fact that global AI products must make difficult choices about data portability and continuity long before a transaction or regulatory event occurs.
Independent operation resets the company’s strategic position
Returning to independent status may give Manus greater flexibility over product direction, partnerships, and market positioning. It may also bring added pressure. A standalone AI-agent company must finance its own infrastructure, recruitment, product development, and customer support while competing with much larger platform companies that can bundle agents into existing ecosystems.
The company’s notice suggests it wants to reassure users that product iteration will continue. It says it is preparing new features and will keep working on general AI agents. Such statements are naturally forward-looking rather than evidence of a specific product roadmap. Still, the focus on continuity is understandable. AI-agent customers are unlikely to commit sensitive work to a service if they fear that access or data management will change without notice.
The separation also returns attention to Manus’ core market position. The company is part of a fast-growing group of firms trying to make AI systems useful for more than conversation. That market will be shaped by execution, reliability, cost, and the ability to integrate with real work. But it will also be shaped by whether users trust providers to protect and preserve their data across corporate changes.
For Manus, the immediate test is operational. The backup window, deletion process, and restoration portal must work as described. The longer test is strategic. Independence can create room for sharper product focus, but it also removes the shelter of a larger corporate structure. In a cross-border AI market, returning to independence is not simply a change in ownership. It is a commitment to rebuild trust, infrastructure, and product momentum under a more complicated global regulatory reality.