Bessent Takes a Big Swing with Outbound Investment Controls

Treasury Secretary Scott Bessent unveiled a proposed rule to update US outbound investment regulations on October 22. Rather than conform with the laissez-faire approach to tech controls that has defined much of Trump 2.0, Bessent, who has been playing a critical role in shaping AI policy, is leaning into outbound investment restrictions with uncharacteristically broad prohibitions. The Treasury update marks a critical signpost confirming our expectation that the high-stakes nature of tech competition with China will impose hard limits on Trump transactionalism. This note breaks down the biggest changes Treasury is proposing, what’s driving those moves, the layers of questions companies will need to navigate to assess coverage, and what lies ahead for an already shaky US-China truce as Beijing builds up its own outbound restrictions in parallel.

On October 22, the Treasury Department released its proposed rule to implement the Comprehensive Outbound Investment National Security (COINS) Act, which Congress passed in December as part of the FY2026 NDAA. The COINS Act, which was catalyzed by the House Select Committee on China, has been a forcing function for the US administration to review and update one of the boldest tech control measures that came out of the Biden era. The proposed rule significantly expands the scope of the outbound program and should give serious pause to any company that has interpreted White House experimentation in deal-making with Beijing as room to hedge and forge partnerships with Chinese counterparts in AI, advanced robotics, chips, and biotech.

The proposal replaces the Outbound Investment Security Program regulations in force since January 2025 (Part 850) with a new Part 851. Comments are due December 7, and the statute requires a final rule by March 13, 2027.

This note breaks down what changed from Part 850 to the new Part 851, where the scope expands most and why, how far the rule reaches into companies that are not headquartered in the United States, and how the rulemaking intersects with Beijing’s own outbound controls.

• Treasury is expanding the yard and raising the fence. The two biggest expansions of note are a flat prohibition on investments in any AI system and on a broad swath of biotechnology-related fields.
• AI compute thresholds have been thrown out in favor of a flat prohibition on AI systems. The prior Part 850 version tried to regulate outbound investment in AI by end use and training compute. The proposal drops these metrics in recognition that the technology is advancing far too fast for a single benchmark to distinguish advanced models. It will be important to watch whether Commerce BIS, especially under potential new leadership post-midterms, adopts the same approach in adapting AI export controls. The prior rule critically left AI “control of robotic systems” undefined. The new proposed rule now broadly defines an AI (rebranded as “super intelligence,” or SI) system to include anything that runs on a model: agent software, humanoid and industrial robots, autonomous vehicles, and drones.
• Biotech made it in, as expected, but in a much bigger way. Prior reluctance to include biotech in the scope of US tech controls was often justified on the grounds that the administration did not know how to define sensitive investments or scope precise technological “chokepoint” parameters, as it has done with semiconductors. As such, we expected that Treasury would first include biotech as a notifiable technology to build up its understanding of investment risks, and at most prohibit some niche segments with clear dual-use potential. Instead, Treasury went big off the bat, with proposed prohibitions spanning biological systems used to produce pharmaceutical compounds, biofuels, and certain controlled agents, toxins, and materials, as well as sequencing technology, bioinformatics tools, laboratory automation, and brain-computer interfaces, mostly without setting technical parameters.
• Semiconductor restrictions have expanded. DUV inputs, high-bandwidth memory, and radiation-hardened chips are prohibited, and compound-semiconductor coverage widens from gallium fabrication to SiC, GaN, and III-V development. AI data center hardware is also in scope: photonic ICs, optical components, and co-packaged optics are prohibited, while HPC interconnects are notifiable. Notification is also required for development or production of any other chip, including test and assembly, and for back-end, test, and mature-node equipment.
• The long arm has extended. The proposed rule would require US person executives at non-US companies to personally notify Treasury of any notifiable deal they “knowingly” direct. Under the prior rule, they only had to stay out of directing deals that would be prohibited. This added layer will likely complicate the recusal process. This will be important in the biotech front in particular: Treasury is concerned about an unlevel playing field where US big pharma is restricted from tapping into Chinese innovation while European and other foreign firms are unrestricted. Concerns that the rules cast US persons as liabilities, create an unlevel playing field, and expose the US to indirect dependencies will likely drive the US administration to pressure trading partners into harmonizing investment screening and export control measures, leveraging broad economic security provisions on export control and investment screening alignment in reciprocal trade deals.
• The scope of covered foreign persons is broadened. The proposed rules broaden the restrictions to cover any entity subject to the “control or direction” of a Chinese person: for example, if a Chinese government body, entity, citizen, or permanent resident can direct its management “whether through the ownership of voting securities, by contract, or otherwise.” Additionally, the definition of “principal place of business” is expanded to include the location of an entity’s main business activities (irrespective of formal HQ). This is likely meant to prevent a Manus 2.0 situation where Treasury struggled to justify its review of US Benchmark Capital’s investment into Chinese start-up Manus AI after it redomiciled to Singapore. The new rule would also extend scope to a large number of Chinese biotech firms registered in the Caymans.
• The 1260H list is a sharp hook. Any notifiable deal becomes prohibited if the target, or any of its affiliates or subsidiaries, appears on the newly added Pentagon 1260H list and forthcoming BIOSECURE list, along with Commerce’s BIS Entity List and Military End-User List and Treasury’s SDN and NS-CMIC list. The 1260H list already includes over 188 entities and is the only entity list that has seen meaningful expansion during Trump 2.0. The low procedural bar for new listings on the 1260H list could give more bite to the rules. It also reflects the increasing relevance of the 1260H list and intent to harmonize various black and red flag red lists (see August 2024, “The Urge to Merge: Streamlining US Sanction Lists Targeting China”).
• Clawback is preserved. The proposed rules preserve the existing exception for deals where US investors buy out all covered foreign holders of an entity or assets located outside a country of concern. This exception will undoubtedly aggravate Beijing’s anxiety around technology flight, as it could shift US incentives away from financing or partnering with Chinese firms through capital injections, JVs, or licensing, toward full buyouts if restrictions prove too cumbersome. (see March 2025, “The Clawback: Reclaiming Strategic Assets from China”).
• Beijing is building the mirror image. China’s Ministry of Commerce (MOFCOM) and Ministry of Science and Technology’s (MOST) proposed updates to the technology export catalog require government vetting before allowing ownership transfers of Chinese AI technology and biotech platforms, while still allowing most out-licensing and outsourced services. While the US is tightening restrictions on outbound technology investments, Beijing is making its own moves to prevent Chinese tech flight, undermine foreign de-risking efforts, insulate its own tech self-sufficiency innovations, and prevent clawbacks of Chinese-owned assets.
• Friction in the US-China relationship will intensify heading into next year. The comment period runs through the APEC summit in Shenzhen and the G20 in Miami, past the November 10 expiry of the tariff truce and the same-day snapback of the BIS Affiliates Rule. Beijing is already holding out on extending the truce until it can get assurance from the US that a slate of tech controls, including a package of measures in Congress, will be neutralized. The White House can get past the summits arguing that the outbound investment rule expansion is not yet a final rule, and the comment period will likely result in some adjustments, particularly in biopharma. Beijing will push back, all while advancing its own outbound regulations in parallel, but will likely be unable to veto the US implementing tighter outbound investment rules around a larger set of technologies.

Bessent goes big

Treasury inherited the COINS Act from Congress, which added two sectors (high-performance computing and hypersonic systems), five countries of concern (Cuba, Iran, North Korea, Russia, and Venezuela under Maduro), and a list of financial sector exceptions for underwriting, bank lending, custody, and payments. It also authorized the secretary to add technology categories that “enable the military, intelligence, surveillance, or cyber-enabled capabilities” of a country of concern and to set their technical parameters.

Treasury used that authority in two big ways: adding biotech and dropping technical parameters for AI, which in effect prohibits the whole category. The discussion on AI in the preamble is clear on why the administration no longer trusts compute thresholds:

This marks the first US regulatory text to abandon training compute as the control variable. We anticipated this shift given how fast AI development is moving and the inability of regulations to keep pace (see August 2026, “Fighting Hubris in AI Strategy: A Layer-by-Layer Dissection of AI Tech Stack Competition”). It will be important to watch whether Commerce’s Bureau of Industry and Security adopts Treasury’s reasoning that inference workloads are a key driver for AI capability to broaden existing AI export controls.

Technology scope expands

The biggest change in the proposed rule concerns the technology lists. Below we break down the changes for AI systems (including robotics), semiconductors, and biotech.

In 2023, the framing of the original outbound rules showed a visible struggle over how to draw lines on emerging AI applications. There was an open question of whether Treasury would restrict AI “exclusively” or “primarily” used for military ends. The 2024 final rule settled on prohibiting only AI designed “exclusively” for, or intended for, military, intelligence, or mass-surveillance end uses, leaving other military-capable AI merely notifiable.

In 2026, under Trump 2.0, the distinction disappears. Embodied AI, agentic software, and any commercial application built on a third-party model would be prohibited investments. Treasury includes a cost analysis, citing PitchBook, that estimates roughly 230 US investors made some 300 deals in the covered sectors in China over 2023–2025, and estimates about 100 potentially covered transactions a year going forward. The broader AI scoping in the proposed rule would shift a number of those notifiable transactions to prohibited. Treasury has already set an example: Its first enforcement action under Part 850, issued in July and announced Oct. 7, imposed a $200,000 penalty against Amidi LLC for an unreported $92,478 investment by its Chinese fund in an embodied-AI startup.

Quantum keeps existing prohibitions and adds design, simulation, and control software and any algorithm or application for quantum information science. Hypersonics are prohibited at the system, propulsion, materials, test-facility, and digital-twin level and notifiable for sensors, guidance, navigation, modeling software and materials “capable of use” in a hypersonic system. This “capable of use” definition could capture a broad swath of components.

Biotechnology

Although not mentioned in the COINS Act, Treasury proactively added biotech as a technology list and placed every covered activity in the prohibited tier. Treasury frames the theory of harm around bioweapons proliferation, brain-computer interfaces, supply-chain dependence, and Chinese indigenization of dual-use capability. Besides medical biotech, the restrictions also reach into agricultural biotech and industrial biotech applications.

The biotech coverage is broad. Prohibition would apply to investments in “biological systems” (left undefined) used to produce, modify, or control certain export-controlled agents or toxins, energetic materials, high-performance materials and biofuels, and, most strikingly, any “pharmaceutical compound.” Coverage would apply to “small molecules, polymers, vaccines, blood and blood components, allergenics, cell and gene therapies, and monoclonal antibodies or other recombinant therapeutic proteins, as well as their delivery mechanisms,” with no therapeutic or technical parameters specified for now. As drafted, it could capture almost any drug-development activity, thereby restricting VC investment in Chinese biotech firms, US pharma firms’ greenfield investment in the country and potentially some of their recent contractual licensing arrangements with Chinese firms. The broad scope will attract intense scrutiny from the pharma industry. Treasury is expecting this and explicitly solicited comment on whether the definition should be narrowed and scoped more precisely to advanced or frontier technologies.

Beyond biological systems, Treasury’s proposed prohibition also reaches biotech hardware and software. As expected, the rule covers flow cytometers, cell sorters and liquid chromatography-mass spectrometry instruments, mirroring the January 2025 Biden-era export controls. The controls also extend to sequencing technology (no technical parameter defined), computational and bioinformatics tools, and any hardware, robotics, or software for automated scientific discovery or experimentation.

Watch the “newco” and “co-co” showdown. Big pharma’s out-licensing deals with Chinese firms have been driving biotech-related headlines and anxiety in Washington. Chinese originators accounted for 40% of global upfront licensing payments so far in 2026 (up from 17% in 2017), and by some estimates Chinese-originated assets could account for at least 35% of FDA approvals by 2040. Because the outbound rules apply principally to equity investments, they would spare strictly hands-off out-licensing deals and fee-for-service drug R&D and manufacturing collaborations, which have been the mainstay of cross-border deals with Chinese firms. But it may be too soon for big pharma to breathe a sigh of relief, as Treasury clarifies that “joint ventures” (which are a covered type of transaction) do not require “the establishment or formation of a new entity” and cover anything “involving the contribution of capital and/or assets by two parties and the sharing of profits and losses.” As such, this expansive definition could potentially capture the “newco” and “co-co” structures behind a growing share of recent headline deals.

Indeed, as early-stage Chinese-originated assets prove their clinical worth and Chinese biotechs hone their lead in new modalities, out-licensing deals have grown more complex, with Chinese firms demanding more from foreign partners. Pfizer’s 2025 in-licensing of 3SBio’s PD-1xVEGF bispecific antibody, for instance, came with a $100 million equity injection. Deals are also converging toward “newco” structures (where the ex-China rights to a Chinese-originated asset are spun out into a separate entity, usually US-based and backed by Western VCs) and co-development and co-commercialization (“co-co”) structures (where the Chinese licensor stays hands-on, sharing clinical trial and development costs and splitting ex-China profits and losses with its foreign partner rather than accepting fixed royalties and milestone payments). These structures remain a minority for now but are growing, especially as Chinese biotechs look to learn from Western big pharma to build their commercialization muscle and ultimately to penetrate international markets on their own.

Mind the European gap. The rules currently apply only to US persons and their controlled foreign entities. European-headquartered pharma groups—including some of the largest by revenue and pipeline, such as Sanofi, Novartis, and AstraZeneca—are not directly bound, beyond any US persons in their executive ranks. As such, some US investors have warned of “eurowashing,” whereby European firms would enjoy unfettered access to competitive Chinese assets and advance them through global clinical trials before filing for FDA approval. We will be watching whether the industry leans into this narrative of unfair restrictions on US firms, and whether the final rule extends the controls’ extraterritorial reach. Short of that, the key question is how Washington uses trade deals and Section 232 agreements to pull foreign firms into compliance.

Assessing coverage under the expanded rule

In 2023, we flagged that the combination of broad definitions for US persons, covered foreign persons, and covered transactions could extend US jurisdiction into transactions by multinationals without a US parent (see August 2023, “Big Strides in a Small Yard: The New US Outbound Investment Regime”). Companies will have to go through a layered checklist to assess whether that’s the case, as we illustrate below in Table 4. Here we run down the changes in Part 851 and illustrate how they could apply to a multinational with China exposure. The outbound restrictions are also likely one of several future steps: If the US intent is to reduce the risk of capital and know-how going to Chinese firms in sensitive technological domains, then it will presumably follow up these rules with pressure on partner countries to harmonize investment screening measures. Investment screening and export control alignment is already written into many of the US reciprocal trade deals, but getting other countries to devise comparable measures will be an uphill battle.

The basic mechanics of the outbound program remain intact. A US person (or a foreign entity it controls) seeking to acquire equity or a convertible interest in a covered foreign person, lend to one on terms carrying equity-like rights, enter a joint venture with one (no new entity required), build or expand operations in a country of concern that will engage in a covered technology, acquire an LP stake in a non-US fund likely to invest in one, or knowingly direct a foreign person to do any of the above must refrain if the technology is prohibited, or notify Treasury within 30 days of completion if it is notifiable.

Companies with a US and China footprint would need to run potential transactions through several layers of questioning. Below is our understanding of how this could apply with the proposed changes in Part 851. Companies should consult with legal counsel to assess direct exposure.

How the proposed rule could apply to companies with a China footprint

In this example, we consider two industrial-technology groups with identical China operations. Company A is incorporated and headquartered in the United States. Company B is a Swiss parent with a US citizen CEO, US citizen board members, and a Delaware subsidiary that runs a lab in Boston. Its China subsidiaries are held directly by the Swiss parent. For purely illustrative purposes, imagine both companies have a broad portfolio covering several technology segments: Both companies have a wholly owned subsidiary in China building industrial robots, a co-development and co-commercialization deal with a Chinese biotech, an IP license to a Chinese chip design firm, and a planned minority stake in a Chinese legacy chipmaker. Here is how the new Treasury rules could impact their operations:

A non-US parent can still trigger coverage through its personnel. Consider a Swiss parent whose Chinese subsidiary takes a minority stake in a Chinese developer of high-bandwidth, low-latency interconnects for high-performance computing, which is now a notifiable technology (and could be prohibited if tied to supercomputers, or if the target or an affiliate is on a covered list). Neither the Swiss parent nor its Chinese subsidiary is a US person, so neither has a filing duty. But the company would have to check for the following exposure: a US person executive or director, including one based at the Chinese subsidiary, who knowingly directs the deal now has a personal filing duty unless they recuse. Any decision made in a US office would be caught regardless of the decision-maker’s nationality. The Swiss parent could also become a covered foreign person itself if its main business activities are in China or most of its financials come from China subsidiaries in a covered technology.

The target test widened in the new proposed rule. An entity anywhere is a covered foreign person if a Chinese government body, entity, citizen or permanent resident can direct its management “whether through the ownership of voting securities, by contract, or otherwise.” Part 850 applied a control test only to government control. The proposed rule applies it to private control with no ownership threshold.

Importantly, “develop” now includes “high-level oversight, direction, approval, conceptual input, or strategic planning,” so a holding company that supervises affiliate R&D would be engaged in the technology. The definition also now covers transforming “design data, blueprints, schema, or other plans into a product.”

A listed affiliate or subsidiary of a target would turn a notifiable deal into a prohibited transaction. The list coverage now includes the Pentagon’s 1260H list and the forthcoming BIOSECURE list, in addition to BIS Entity and Military End-User List and Treasury SDN and NS-CMIC lists. The FCC Covered List is not yet included in this harmonization list.

The rule does not bar US nationals from working at Chinese companies, does not cover sales, licensing or procurement, and does not cover transactions completed before the final rule’s effective date. It also does not grandfather deals signed but not closed by the effective date.

Another big stressor to the Busan truce

Beijing will no doubt be alarmed by the first big swing by Treasury under Trump 2.0. The outbound rule expansion, on top of building FCC import restrictions, a slate of pending export controls in Congress, and a potential change in guard at Commerce post-midterms are pointing to the limits of Beijing’s strategy to leverage its control over critical inputs to neutralize US tech controls.

The final rule will not come until mid-March, past the Trump-Xi leadership meetings planned in mid November and mid-December. Bessent, who has been steering the US-China bilateral relationship, could assert that the outbound investment restrictions are still being drafted and that they are not up for negotiation. We will be watching for Beijing to assert demands on softening the measures, especially in covering Chinese biopharma firms. Given how broadly scoped the biotech section is in the current draft, Bessent could use expected rollbacks to manage a negotiation with Beijing and show where the US is willing to pare back in return for Beijing demonstrating comparable restraint, particularly in controlling supply of critical inputs. With the US now actively leaning into biotech controls, Beijing faces a difficult decision over whether it wants to risk the political backlash over disrupting healthcare supply chains in a retaliatory response.

Beijing, meanwhile, is building outbound defenses of its own. MOFCOM issued a draft rule on regulating Chinese trade secrets that would have the ministry investigate foreign firms whose infringement threatens China’s “sovereignty, security, or development interests” and respond with curbs on their China trade and investment, fines, and “other necessary measures.”

A parallel MOFCOM–MOST draft revision of the technology export catalogue would add frontier AI models and agents, AI chip, DRAM and 3D NAND design, AI cluster technology, cell-and-gene-therapy and siRNA platforms, and next-generation solar manufacturing. Ironically, its AI criteria keep the US-defined compute thresholds (10^26 FLOPs, 2 trillion parameters) while adding two alternative triggers: a global top-10 ranking or training on “core” or “important” data. Any ownership or control change that moves Chinese technology under an overseas entity would trigger registration. Beijing has several strategic objectives at play: undermine Western de-risking efforts that threaten China’s own industrial base, prevent Chinese tech flight, insulate its own innovations as it pushes self-sufficiency in semiconductors and AI, and prevent western clawbacks of Chinese-owned entities. Both sides appear to be converging on allowing licensing, but barriers to IP transfer remain a risk.

Finally, US trading partners can now expect additional pressure as the US administration tries to make the economic security provisions it negotiated into reciprocal trade deals mean something when it comes to aligning on investment screening and export controls. Bessent is likely well aware of arguments that these rules cast US persons as liabilities, create an unlevel playing field, and expose the US to indirect dependencies when other major markets have unfettered access to Chinese tech firms. We will be watching for just how far the administration takes the pressure campaign, especially given the potential for a change of guard at Commerce leadership and BIS post-midterms (see October 29, “Trump 2.0 in a Divided Government”).