Outbound Investment Rules: What Sensitive-Tech Investors Must Disclose

For two decades the national-security conversation about cross-border capital in the United States concerned inbound investment, reviewed by CFIUS. Outbound investment is now the second gate. The shift is easy to describe as reverse CFIUS, and the shorthand is useful, but it hides a difference: the outbound regime is largely a rules-based prohibition and notification system rather than a case-by-case review, so the burden of getting the classification right falls on the investor.

The programme rests on an executive order of 9 August 2023 declaring a national emergency. Treasury issued the final rule on 28 October 2024, and the rule took effect on 2 January 2025, according to Morrison Foerster's review of the text. Treasury's own programme page confirms that notifications are filed through its Outbound Notification System. The rulemaking was not rushed: Treasury received 60 comment letters on its advance notice and over 40 on the proposed rule, which helped shape the more detailed final text.

For a head of corporate development the implication is that outbound investment is no longer a policy debate to monitor but a compliance obligation that attaches to individual deals. Strategic minority stakes, joint ventures with foreign partners and venture co-investments can each fall within scope depending on the target's activities and ownership. The first question is no longer whether the deal is commercially attractive but whether it is a transaction the rule reaches at all.

Add an outbound question to the deal intake form this quarter if it is not already there. It should ask where the target is organised and operated, what technology it develops or produces, who owns or controls it and whether the investment gives rights beyond a passive stake. Route any yes or uncertain answer to counsel before diligence spend is committed. Treat this memo as information, not legal advice, and involve qualified counsel on classification.

It is worth being clear about what the outbound rules are not. They are not an approval process in which an investor submits a proposal and waits for a decision, and they do not generally give the government a veto over a deal that falls outside the defined categories. That makes the regime faster than a review, but also less forgiving. There is no clearance to rely on, so the investor's own file of reasoning is the only protection if the classification is later challenged.

A common misreading is that the rules chill all investment in China. They do not. The programme is designed around a short list of technologies where the stated concern is military, intelligence or surveillance advantage, and a defined set of foreign persons. That narrowness is good news for most corporates, because the great majority of cross-border deals will fall outside it, but it also means the cases that do fall inside leave little room for argument.

Treasury's programme page states that the order identifies three categories of national security technologies and products: semiconductors and microelectronics, quantum information technologies, and artificial intelligence. The Federal Register notice records that the President identified one country of concern, the People's Republic of China, together with Hong Kong and Macau. Within each category the rule distinguishes between activities that are prohibited outright and those that must be notified to Treasury after completion.

The implication is that classification, not geography alone, decides the outcome. A software investment in a Shanghai start-up is not automatically caught, while a chip-design business with a thin China connection might be. Corporate development teams therefore need technical input at the screening stage, from engineers or product leads who can say what the target actually builds and how it is trained, deployed or fabricated, rather than relying on how the target describes itself in a pitch deck.

Ask the target for a plain-language description of its products and customers, and have your technical lead confirm it against the rule's categories. Record who reviewed it and when. Where the activity sits close to a category line, take a conservative position and seek counsel's view before engaging further. Keep the evidence file with the investment memo so that the reasoning survives staff turnover.

The same facts that decide whether a deal is in scope also tend to determine how it will be seen by banks, insurers and partners. A lender asked to fund an acquisition in a covered sector will want comfort that screening was done. A strategic partner in the United States will ask the same question before sharing intellectual property. Treating the rule as a commercial credential, not only a legal hurdle, helps teams win internal support for the effort involved.

The rules are unusually quantitative. Rather than banning broad categories, they set thresholds that can be tested with documents. This is helpful, since it reduces discretion, but it shifts work to diligence: an investor must obtain evidence on technical specifications, training compute and ownership percentages that a target may not volunteer and that a minority investor may struggle to verify without contractual rights.

Morrison Foerster's analysis reports that the final rule treats as prohibited any AI system trained using computing power greater than 10^25 computational operations, and as notifiable any system trained using more than 10^23. Ownership tests also matter. Under the COINS Act, Arnold & Porter reports, a covered foreign person includes one owned in the aggregate, directly or indirectly, 50 percent or more by covered entities. The limited partner exception, by contrast, applies where the partner's investment in the fund is $2,000,000 or less.

For fund managers the practical effect is that responsibility attaches to the general partner's decisions, while limited partners have their own, narrower position. The exception for smaller passive limited partners helps, but it is conditional, and managers should not assume that a fund's structure insulates the portfolio from the rule. For corporates, the 50 percent test means that indirect ownership through holding companies must be traced to the end, not assessed at the first layer.

Insert contractual information rights into term sheets for any investment with a technology or country nexus: a representation on covered-person status, a covenant to supply technical and ownership information, and a right to unwind or restructure if the facts change. Where the investor holds a minority position and cannot compel disclosure, document the efforts made. A record of reasonable inquiry is a stronger position than silence.

The software industry offers a useful illustration of why the technical review matters. A company that sells analytics tools may use machine learning in its product without developing a model of the kind the rule describes, while another may train its own large model as its core asset. The two can look similar in a pitch and be classified very differently. Corporate development teams are not expected to resolve that alone, but they are expected to ask.

The cost of getting classification wrong is not limited to a fine. A transaction that should have been prohibited, or that should have been notified, is a legal defect in the deal. Counterparties may demand renegotiation, lenders may refuse to fund, and the investor may be asked to unwind. The monetary penalty is real, but the larger exposure is usually the disruption to the deal and the reputational effect on future relationships with co-investors.

Morrison Foerster reports that the maximum civil penalty for a violation is the greater of $368,136, adjusted annually for inflation, or twice the value of the underlying violative transaction. Because the second limb scales with the deal, a large investment carries a large ceiling, and the first limb ensures that small deals are not unpoliced. Treasury's programme page cites the International Emergency Economic Powers Act as the authority for penalties and remedies, which carries both civil and criminal consequences.

The implication for a corporate development leader is that outbound risk should be priced into the deal timetable and the purchase agreement as early as signing. A condition precedent that depends on a regulatory determination, or an indemnity for misclassification, may be worth negotiating when the target is in or near a covered sector. These protections are cheaper to obtain before exclusivity ends than after, and they signal seriousness to the counterparty.

Treat late discovery as a scenario in the deal model: what happens to timing, financing and valuation if a covered classification emerges in the final month? Run it for any transaction in or adjacent to the three technology classes. If the downside is large, move screening further left in the process. The cost of an earlier, lighter review is trivial compared with that of a stalled closing.

Boards and investment committees should expect to see this analysis summarised in the paper for any deal with a technology nexus. A single paragraph is enough: the classification reached, who reached it, what evidence supports it and what residual uncertainty remains. This discipline has a side benefit. It teaches deal teams to notice the early signals of an in-scope transaction, and over time the number of late surprises falls, because the questions become habitual.

Congress has now put the programme into statute. The Comprehensive Outbound Investment National Security Act, referred to as the COINS Act, was enacted through the National Defense Authorization Act, and it does more than codify Treasury's rule. It broadens the list of countries of concern, changes how covered foreign persons are defined and directs Treasury to publish more of the working detail, which will change the compliance picture even before the new regulations arrive.

Arnold & Porter reports that the President signed the Act on 18 December 2025 and that Treasury must issue implementing regulations within 450 days of enactment, which we calculate as around mid-March 2027. It retains a duty for US persons to notify Treasury within 30 days of completion and authorises $150 million annually for FY26 and FY27 for Treasury. White & Case reports that countries of concern will include Cuba, Iran, North Korea, Russia and Venezuela, and that hypersonic systems and high-performance and supercomputing are likely to be added as technology areas.

Two consequences follow. First, the compliance perimeter is not fixed: until the regulations land, investors must plan for a rule that may be wider than the current one. Second, White & Case notes Treasury must create a non-exhaustive public database of covered foreign persons, which could simplify diligence for the entities listed but will not replace it, since the list is non-exhaustive and the control-based definitions reach beyond named entities.

Calendar the 450-day deadline and designate someone to track Treasury's rulemaking, comments and guidance. In the interim, apply the stricter reading where the two regimes differ, particularly on ownership and control. Review existing portfolio holdings in hypersonics and supercomputing now, since early identification gives time to restructure or seek advice. The cheapest moment to learn about a gap is before it becomes a reporting obligation.

Corporates face a particular version of this problem with minority stakes taken for strategic reasons. A venture arm that invests alongside a financial sponsor in a round may not control any decision, but it still makes an investment, and the sponsor's co-investors may include parties with a country-of-concern link. The corporate should ask the lead investor for a syndicate list before committing, and should write the answer down. It takes little time and may avoid a dispute months later.

The rules did not create the retreat of US capital from Chinese technology; they formalised a trend already under way. Understanding the pattern helps a corporate development team judge where requests are most likely to arise, because the sectors most exposed to the rules are the same ones in which deal flow persisted longest and where advisers and co-investors will keep bringing proposals.

S&P Global Market Intelligence found that US venture-backed funding rounds in China totalled $3.93 billion across 116 deals in 2023, the lowest value and volume since 2013. An Atlantic Council testimony states that US outbound investment to China in 2023 was 30 percent of its 2021 value. PitchBook reports that deal value in China with US investor participation peaked in 2018 at a record $48.5 billion, and that nondomestic investors fell from about 9 to 10 percent of China AI deal value at their peak to 3.7 percent in 2023, recovering to 7.1 percent in 2025.

The sector mix explains why the rules matter now. PitchBook reports that AI represented 20.5 percent of China venture deal value in 2024, up from 11.7 percent in 2023, and 19.6 percent in 2025. In 2025 it recorded over $4.2 billion of AI software deal value, $2.7 billion in AI hardware and $1.6 billion in AI semiconductors. These are vendor data with overlapping categories, but the direction is clear: capital has re-concentrated in precisely the technologies Treasury listed.

Use the data as a prompt, not a conclusion. Ask whether your pipeline includes any co-investment, secondary or fund-of-funds position with China AI or chip exposure, and whether intermediaries who bring you deals are screening for the rules. Brief deal teams that a request to join a round alongside Chinese or Hong Kong investors in an AI or semiconductor company requires an outbound check before any substantive engagement.

Readers should also keep the data in perspective. The PitchBook and S&P series are commercial datasets with their own definitions, and they capture venture and private equity activity rather than all forms of outbound capital. They do not show how much of the retreat was caused by regulation, by the weaker Chinese economy or by investor caution. Their value here is directional: they identify the technologies where the regime is most likely to be tested in practice, which is where screening effort should be concentrated.

Most organisations that struggle with a new regime do so because they place the control in the wrong spot. Legal reviews the agreement near signing, discovers a classification issue and has limited ability to change the deal. The remedy is structural: a short, standard triage at intake, owned by corporate development, with escalation rules that bring counsel in early for the small share of deals that need them.

The triage need not be elaborate. It asks five things: the target's location and operations, its technology, its ownership and control, the investor's rights and the participation of other investors. Treasury's definitions, as described by Morrison Foerster and Arnold & Porter, rest on the same facts. A completed form, kept with the investment memo, provides a consistent audit trail and gives the board confidence that the question is being asked of every deal rather than only the obvious ones.

For funds, the same triage should extend to portfolio monitoring. Companies change business lines, raise rounds from new investors and acquire technology, so a deal that was out of scope at signing can move into scope. Build an annual attestation from portfolio companies and a trigger for re-screening when they announce a pivot, a round or a change of control. For corporates, apply it to joint ventures and strategic stakes as well as financial investments.

Start with a thirty-day sprint: draft the intake form, agree escalation criteria with counsel, back-test it on the last year's deals and brief every deal team. Then run a quarterly review of cases escalated and outcomes. We would expect the first back-test to surface a handful of past transactions worth a second look. Better to find them in a controlled review than through an inquiry.

Finally, resist the instinct to wait for the final COINS regulations before acting. The core of the current regime will almost certainly remain: the technology classes, the notification mechanism and the penalty structure. The incremental work to build an intake gate is modest and the benefits accrue immediately. When Treasury's regulations arrive, a team that already has a working process can adapt it in weeks, while one that waited will have to build from nothing under time pressure.

Our M&A & Corporate Development work helps teams design the deal-intake triage, the evidence file and the escalation path that bring outbound screening into the transaction process. We support the process alongside qualified counsel and regulated advisers; this page is information, not legal or investment advice.