The Chinese AI and venture capital community was shocked by Meta’s swift acquisition of Manus, an AI agent company from Wuhan, China, in a deal worth an estimated $2–3 billion. For Meta, this deal is its third-largest acquisition to date. It bets on Manus to win the “agentic” AI race depending on Manus’s autonomous task-execution capabilities marketed as a “digital employee” capable of independently planning and delivering complex workflows.

From “national pride” to “runaway”

Manus made its international debut in March 2025, by unveiling what it touted as the world’s first general-purpose AI agent, an autonomous system that could execute complex tasks such as drafting reports, analyzing stocks, and coding, with minimal prompting. The product went viral and gained massive domestic and international attention. Chinese state media and the local government of Wuhan lauded it as “the next DeepSeek” and showed strong eagerness in supporting this new golden child of AI innovation.

However, in a drastic overturn/contrast, after Silicon Valley venture capital firm Benchmark invested in Manus with 75 million dollars in April, Manus swiftly made decisive moves to segregate itself from the Chinese market. It closed its Wuhan and Beijing offices, deleted its Chinese social media account, fired Chinese employees except for the core team, and moved to Singapore. Users with Chinese IPs were even blocked from accessing its site.

This dramatic shift triggered polarized sentiments in China. On one hand, venture capital and tech communities are excited about an emerging new playbook of accessing the lucrative US capital market. As long as you totally abandon your Chinese identity, you can still be accepted by the US capital market and make money through acquisitions. On the other hand, some commentators criticized the company for “running away” from the Chinese market. From the angle of the Chinese government, the acquisition was especially alarming as it sets a dangerous precedent that domestic innovators could just abscond to the U.S. despite all the support they gained from the domestic talent pool, policy encouragement, and industry advantage. Such grievance is not inconsequential, as Beijing also attempts to review whether the deal is consistent with its export controls (this is our previous article on China’s nascent export control regime).

The Benchmark investment and OISP screening

Why did Manus gain such a sudden lucrative acquisition, and also determine to cut its ties with China so decisively in a short time span of 9 months? The startup defended itself by stating its “DNA of internationalization” from the very beginning, and their retreat from China is just a reflection of their strategy of becoming a truly international company, in order to avoid foreign blockades and utilize American LLM APIs such as Claude. More importantly, after receiving investment from Benchmark, Manus swiftly went on a path of restructuring and relocating in order to navigate through the heavy American investment screening crossfire.

The investment was an audacious move for Benchmark; as US venture capitalists become more national security aware, they have avoided investing in Chinese firms. The deal was subject to the “reverse CFIUS” – the U.S. Outbound Investment Security Program (OISP) rules that impede American firms investing in Chinese technological companies. In hindsight, before the Benchmark investment, Manus apparently falls into the OISP covered category – a China-based company operating in China. Hence, after Benchmark’s investment and guidance, all its relocation and segregation moves serve to conform to OISP; Manus has spent over 9 months to implement an “identity engineering” task to become fully “non-Chinese” and “international”.

How “Chinese” are you?  U.S. government’s expanding approach of determining company identity

From FIRRMA and the restrictions on Huawei and ZTE to the recent OISP rules, the U.S. government has steadily broadened the definition of a “covered entity.” Moving beyond the simple principle of legal nationality, it has adopted a holistic risk-assessment mechanism that scrutinizes inbound and outbound investments, as well as domestic market entries, targeting Chinese-background firms and individuals. This evaluation is based on a wide range of factors, including headquarters location, revenue sources, expenditure destinations, ownership percentages, data storage location, algorithm control, and government connections.

Starting from 2018, the Foreign Investment Risk Review Modernization Act (FIRRMA) allowed U.S. authorities to scrutinize even non-controlling minority stakes by Chinese investors in critical tech sectors. Around the same time, Chinese telecom giants Huawei and ZTE were regarded as entities with strong connections with the Chinese state, and eventually blacklisted in the U.S. network by 2019-2020. During this period, the US regulators mainly focused on companies being under the Chinese jurisdiction or influence.

However, as the TikTok saga shows, moving out of the Chinese jurisdiction wouldn’t guarantee a free pass from the U.S. government. Even though TikTok has long separated its Chinese and international branches, and developed data localization and security plan in the U.S. territory (see Project Texas), the U.S. regulators still pressed TikTok to sell and restructure. The eventual deal, reached in late 2025, forced ByteDance to reduce its stake to 19.9% and cede majority ownership to U.S. investors. U.S. user data, critical algorithms, and content decisions must be controlled by a new U.S.-led joint venture (see our previous post). This case shows what the U.S. really wants over a China-originated company when it carries strategic value: to gain absolute controlling power over it.

Recent developments of the Comprehensive Outbound Investment National Security (COINS) Act further exemplifies the intentionally expansive nature of the U.S. government’s definition of a Chinese company. The COINS Act broadens the scope of “covered foreign persons” by replacing the OISP’s fixed equity and voting thresholds with a qualitative “direction or control” standard, which captures entities regardless of their specific ownership percentages. This shift ensures that minority-owned or non-technical Chinese entities still remain subject to OISP.

Pushing this current logic and trend further, an extreme but plausible situation would be: whenever a China-originated company is deemed to carry high strategic value (be it in AI, telecommunications, social media, etc.), even entities that fall outside this strict definition (e.g. headquartered offshore with mixed ownership) can be treated with suspicion if they have Chinese DNA in their founding team or technology.

Mind the gap: the stack of Identity engineering

By summarizing the current U.S. regulations and political climate, a new Chinese AI firm will have to painstakingly conduct a whole stack of identity engineering if they are still seriously seeking to obey U.S. regulations, enter the U.S. market, and/or receive U.S. investment:

• Incorporation & HQ abroad: register and headquarter outside mainland China, Hong Kong, and Macau SARs.
• No majority China exposure: ensure the majority of its revenue and expenditure are not primarily tied to the Chinese market.
• Avoid state linkage: decline investments from state-owned funds and avoid state contracts and affiliations.
• Isolate data flows: store user and operational data outside China and block any cross-border transfers back to Chinese servers.
• Rebrand globally: De-emphasize Chinese origin in marketing, press, and investor materials.

In a more extreme playbook, when more restrictive and racist requirements emerge from the U.S., they might even have to:

• Be led by non-Chinese management and staff: limit Chinese passport holders in executive, technical, or board-level roles.
• Have zero China operations: shut down domestic business entirely and block Chinese IP access to platforms.
• Supply chain decoupling: move manufacturing and hosting relationships out of China.

If a China-originated company truly adheres to all the requirements here, can they still be defined as Chinese companies? Manus made a clear choice to totally abandon its home market identity, but not all firms are audacious enough to burn so many bridges. This level of compliance with U.S. regulations is, in effect, a clear act of “taking a side”, which will inevitably carry consequences from the Chinese regulator and market.

Conclusion

Although some commentators may argue that identity engineering is just companies’ reactive measures to bypass and dodge U.S. regulation, such policies still pose long-term costs to Chinese companies.

Chinese technology firms like Manus essentially attempt to leverage the comparative advantages of both China and the U.S. On the one hand, they thrive from the high-quality talent pool, advantageous application innovation environment, and strong manufacturing capacity from China. On the other hand, they still want to enter the US market due to its high liquidity, strong consumption power, and access to computing power. Under the current restrictive and discriminative policies from the U.S. government, they must implement a full stack of “identity engineering”. However, when the costs for compliance have reached a tipping point where they exceed the gains, they may need to consider another path.

More importantly, the pivotal question remains for some Chinese firms: despite all their efforts to rebrand, restructure, and even alienate their home market, will the U.S. regulators ever truly accept them as “non-Chinese”? I doubt so, given the contingent nature of U.S. investment policies amid strategic competition with China, the final determination of “who they are” is hardly theirs to make.

The post Identity Engineering: Why a leading Chinese AI startup abandons its home market appeared first on Internet Governance Project.