Manus has given affected users until 7:59 a.m. Singapore time on Aug. 23 to back up records created from Dec. 29, 2025, as the AI-agent developer completes its separation from Meta and returns to independent operations. The company said it will delete certain data produced during that period to comply with regulatory requirements tied to the unwinding of Meta’s proposed acquisition.
The service will undergo a temporary interruption from Aug. 23 through Aug. 24 while the deletions are carried out. Manus said access for affected accounts will be suspended during the process, with service scheduled to resume at 8:00 a.m. Singapore time on Aug. 25. Users will then be able to retrieve backed-up information through the company’s backup tool.
Manus said the data-handling process is unrelated to a data leak or cybersecurity incident. Users whose accounts are not affected do not need to take action, according to the company’s Aug. 11 notice.
Backup window covers new task records
The company has allowed users to make multiple backups before the deadline. That matters for users continuing to run tasks during the transition: a backup only captures records available at the time it is created, meaning task data generated afterward requires another backup to preserve the latest version.
Manus has not framed the process as a full deletion of all user information. Its notice specifically focuses on certain data generated on or after Dec. 29, 2025, the date Meta first announced its plan to acquire the company.
After Manus resumes standalone operations, user data will continue to be stored in the United States and Singapore, the company said. The storage arrangement places the operational split largely around handling data created during the attempted ownership period rather than moving the platform’s entire data footprint to a new jurisdiction.
For users, the immediate practical issue is whether any research tasks, generated reports, code projects, processed files, or other agent outputs created after Dec. 29 are needed after the Aug. 23 cutoff. Manus describes its product as a general AI-agent system that can take natural-language instructions and perform multi-step work including information retrieval, report writing, file processing and software development.
China ordered Meta deal to be unwound
The disruption follows a prohibition decision issued on April 28 by China’s foreign investment security review mechanism office, which operates under the National Development and Reform Commission. CCTV News reported that the authority barred Meta’s acquisition of Manus and required the parties to revoke the transaction.
Meta announced its acquisition plan on Dec. 29, 2025. Financial terms were not disclosed by the companies. Earlier reports put the transaction at roughly $2 billion, with employee retention arrangements potentially lifting its total value to $2.5 billion.
Manus said the companies subsequently began separating their operations and stopped sharing data. That operational separation was completed in May, while the current backup and deletion schedule represents the final data-handling stage before the company resumes work independently on Aug. 25.
The ruling demonstrates the reach of China’s foreign investment review system over an AI company whose corporate base had moved to Singapore but whose origins and engineering history remained closely linked to China. Manus was developed by Butterfly Effect, a company founded in 2022 that began in China before relocating to Singapore.
A fast-growing agent platform returns to standalone status
Manus entered the public market on March 6, 2025, after its team had previously built Monica, an AI assistant integrating multiple large-language-model capabilities. The later Manus platform expanded that work from chat-based assistance into task execution, using virtual computing environments to carry out work across multiple steps.
The company said in December 2025 that it had exceeded $100 million in annual recurring revenue after eight months on the market, with its total revenue run rate reaching $125 million. Manus also said its systems had processed more than 147 trillion tokens and created more than 80 million virtual computers.
Those figures illustrated why the company became a target for Meta: agent platforms are trying to turn language models into systems that can complete practical work rather than merely answer prompts. The forced reversal leaves Manus operating independently at a point when the cost of computing infrastructure, model access and global data operations remains high.
Manus has not publicly disclosed its future ownership structure or a new financing plan. Prior reporting indicated that its founding team had explored raising about $1 billion to buy back shares, with early supporters including ZhenFund and Hony Capital participating in discussions.
Founder Xiao Hong leads Butterfly Effect. Before Manus, Xiao created Wuhan Nightingale Technology and developed WeChat-oriented products including Yiban Assistant and Weiban Assistant, which the company said served more than 2 million business users. Co-founder and chief scientist Ji Yichao previously worked on mobile browsing and information-extraction technology through Peak Labs, while co-founder Zhang Tao held product roles at ByteDance and Guangnian Zhiwai.
New Shanghai lab targets long-running agents
The Manus split comes as competition in AI agents is drawing experienced model developers into new ventures. Lin Junyang, formerly the technical lead of Alibaba’s Qwen project, has announced Pragmatik, also known as p7k Labs, in Shanghai.
Lin left his Qwen role on March 3, 2026, after working on large-model development, reasoning, agent training and model interaction with external environments. Pragmatik says it will build digital agents for knowledge work, enterprise operations and industrial processes, alongside physical agents intended to operate in real-world settings.
The lab identifies reasoning, tool use, learning from feedback and coordinated action as the core abilities required for next-generation agents. Its goal is to create systems that can remain active in an environment, alter their approach as conditions change and pursue longer tasks—precisely the capabilities that have made operational data, infrastructure and regulatory control central issues for companies such as Manus.
Explore how AI reshapes trading automation with Toobit’s AI copy trading for smarter, regulation-aware portfolio strategies.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
