White House Alleges Chinese Companies Orchestrating Widespread Artificial Intelligence Theft Campaign

The White House has issued its most direct warning yet that Chinese firms are executing industrial-scale theft of American artificial intelligence, naming distillation campaigns as the mechanism and citing new intelligence that foreign entities are systematically dismantling US competitive advantage in the technology that will define the next generation of economic power.

Michael Kratsios, Director of Science and Technology Policy, wrote in an internal memo that foreign entities, principally based in China, are exploiting American AI companies through coordinated distillation operations. The technique works by running thousands of individual accounts across AI platforms, using those accounts to extract non-public model information through jailbreaking attempts, and feeding that intelligence into rival model development. The White House outlined four responses: sharing intelligence on tactics and actors with US companies, coordinating joint defenses, developing best practices for detection and remediation, and exploring accountability mechanisms for foreign actors. No enforcement timelines or specific sanctions were disclosed.

Anthropic has already named three Chinese AI laboratories, DeepSeek, Moonshot, and MiniMax, as the firms conducting distillation attacks against its models. OpenAI has separately accused DeepSeek of copying its technology. Neither DeepSeek, Moonshot, nor MiniMax responded to requests for comment from the BBC. A representative of China's US embassy rejected the framing, stating that China's development "is the result of its own dedication and effort as well as international cooperation that delivers mutual benefits," and pushed back against what it called "unjustified suppression of Chinese companies by the US" [1].

For institutional capital, this is not an IP enforcement story. It is a structural repricing event across the AI investment landscape. The memo lands at the precise moment DeepSeek has released its most competitive models to date, Brent crude sits above $106 per barrel, and the global technology supply chain is already under stress from the Iran conflict's disruption of Strait of Hormuz transit [2]. The convergence of these forces creates a portfolio risk that most PE and growth equity models have not yet priced.

DeepSeek-V4-Pro Arrives as the White House Fires Its Warning Shot

The timing is not incidental. On April 24, 2026, DeepSeek launched preview versions of DeepSeek-V4-Pro and DeepSeek-V4-Flash, one year after its original model shook Silicon Valley. The Hangzhou-based startup claims DeepSeek-V4-Pro beats all rival open-source models on mathematics and coding benchmarks, and trails only Google's Gemini 3.1-Pro, a closed model, on world knowledge tasks [5].

DeepSeek described its performance gap against OpenAI's GPT-5.4 and Gemini 3.1-Pro as "marginally short," characterizing the deficit as "approximately 3 to 6 months" behind state-of-the-art frontier models [5]. If accurate, that is a remarkable compression of the capability gap. The original DeepSeek-R1, released in January 2025, reportedly cost less than $6 million to build, a figure that prompted skepticism from analysts who argued the startup likely had access to greater funding and more advanced chips than acknowledged [5].

The White House memo and the DeepSeek product launch are two data points in the same thesis: China is closing the AI gap faster than the capital allocation models of US technology investors assumed, and at least some of that acceleration may have been built on extracted US intellectual property.

The Distillation Economy: How IP Theft Inverts the Unit Economics of AI Investment

The financial logic of AI distillation is precise. If a US company spends hundreds of billions of dollars training a frontier model, and a foreign entity can extract the functional output of that investment through systematic distillation at a fraction of the cost, the return on invested capital for the US company collapses. Kratsios acknowledged this directly, stating that the aim of distillation campaigns is to "systematically undermine American research and development and access proprietary information" [1].

This matters to PE and growth equity investors because AI valuations at the frontier are built on defensibility assumptions. A model that cost $10 billion to train commands premium licensing fees and enterprise contract terms because no competitor can replicate it cheaply. Distillation breaks that assumption. If the effective cost of approximating a frontier model falls toward the single-digit millions, the moat narrows, pricing power compresses, and the multiples at which frontier AI companies have been capitalized require revision.

The irony is that the Kratsios memo, by not naming specific US companies or disclosing enforcement mechanisms, does little to repair the defensibility of US AI IP in the near term. The White House said it will "explore" accountability mechanisms. That language does not protect the training investments already made.

Geopolitical Stack: Hormuz, AI Chips, and the GCC Pivot

The macro context compounds the technology risk. Brent crude crossed $106 per barrel on April 24, 2026, up nearly 5 percent from its Wednesday close, driven by the US-Iran confrontation in the Strait of Hormuz [2]. Only nine commercial vessels transited the strait on Wednesday, compared to an average of 129 transits per day before the US and Israel launched military operations against Iran on February 28 [2].

The strait carries approximately one-fifth of the world's oil and natural gas supply [2]. The Persian Gulf region supplies roughly 50 percent of Europe's jet fuel imports, according to MercoPress [3]. Jet fuel prices have doubled since hostilities began, reaching a record $1,840 per metric ton by early April [3].

For AI infrastructure investors, the Hormuz disruption is a secondary but material input cost shock. Data center construction, hardware logistics, and energy-intensive GPU operations all carry embedded energy costs. A sustained $106 per barrel oil price environment raises the operating cost floor for compute-intensive AI workloads at exactly the moment when US companies are being asked to spend more on model security and IP protection.

The GCC angle is asymmetric. Gulf sovereign wealth funds, including entities from the UAE and Saudi Arabia, have committed capital to US AI infrastructure at scale. A deteriorating US-Iran security environment in the Gulf does not automatically reduce those commitments, but it introduces a geopolitical variable into LP relationships that fund managers with GCC capital bases will need to manage actively.

LATAM Exposure: Supply Chain Fragility and the Open-Source Arbitrage

Latin America is a second-order casualty. The region's technology sector has moved aggressively toward AI-enabled services, with Brazilian, Mexican, and Argentine firms adopting open-source models, including DeepSeek's prior releases, as the cost-efficient foundation for local product development.

The White House memo creates a compliance overhang for LATAM technology companies that have built on Chinese open-source AI. If the US moves toward export controls or access restrictions on firms that have used distillation-derived models, LATAM companies face a forced technology stack review. That is not a 2026 budget item most boards have modeled.

The open-source nature of DeepSeek's new V4 models, meaning developers are free to use and modify the source code freely [5], makes enforcement architecturally complicated. You cannot un-release open-source code. The White House has not disclosed how it intends to address derivatives of models that are already freely distributed. Until that answer arrives, LATAM technology investors carry unquantified compliance exposure.

Investment Positioning: Where the Money Should Move

Three implications follow directly from the source material.

First, US AI infrastructure companies with proprietary model weights face a bifurcated risk. The distillation threat identified by the White House erodes moat assumptions; the absence of enforcement mechanisms leaves that erosion unaddressed. Investors in closed-model AI companies should pressure management on quantified estimates of model leakage and the cost of detection and remediation infrastructure, which the White House has now signaled will become a baseline expectation.

Second, open-source AI plays, including any investment thesis predicated on DeepSeek derivatives or Chinese open-source model stacks, carry escalating regulatory risk. The memo's language around accountability for foreign actors is vague today. It will not remain vague. Position sizing in anything downstream of Chinese AI development should reflect that trajectory.

Third, energy transition and infrastructure assets in the GCC offer a hedge. The Hormuz disruption is creating a structural incentive for Gulf states to accelerate energy diversification and data infrastructure investment. Capital flowing into GCC technology and infrastructure funds benefits from both the oil revenue windfall at $106 per barrel and the geopolitical urgency to build sovereign AI capability before the US-China technology war closes off options.

The Plocamium View

The White House memo is less significant as a policy document than as a signal of where enforcement is heading. The four commitments Kratsios outlined, intelligence sharing, coordination, best practices, and accountability exploration, are the precursors to a harder regulatory posture that will arrive before 2026 is out.

Our thesis: the distillation disclosure will accelerate a structural bifurcation in the global AI market. One cluster will form around US-allied, export-controlled, closed-model AI infrastructure. A second cluster, anchored by Chinese open-source releases from DeepSeek, Moonshot, and MiniMax, will continue to propagate freely across non-aligned markets, including much of LATAM and Southeast Asia. The White House cannot close that second cluster through memo writing.

The second-order implication is where the alpha sits. Anthropic and OpenAI are now publicly documented as victims of distillation campaigns. That creates a political and commercial incentive for them to support US government intervention in ways they have historically avoided. Expect both companies to become more active in lobbying for export controls on compute and model weights. That lobbying, if successful, raises barriers to entry for US-market AI investment and compresses the addressable pool of competitors. That is, paradoxically, bullish for defensible US frontier AI positions in the medium term.

The Iran conflict adds a time dimension that markets are underweighting. At $106 per barrel and nine Hormuz transits per day versus a pre-conflict average of 129, the energy and logistics cost base for global technology deployment is rising structurally [2]. AI training runs are energy-intensive. If energy costs remain elevated while AI IP security costs rise simultaneously, the companies that survive will be those with the deepest capital reserves and the most defensible model architectures. That is a consolidation thesis. In PE terms, it is a vintage opportunity: prices for second-tier AI assets will compress before strategic acquirers move.

Position for consolidation. The distillation memo is the starting gun.

The Bottom Line

The White House has named China's AI distillation campaigns as a systemic threat to US technological leadership, and DeepSeek's same-day release of its most capable models yet confirms the urgency is real. Institutional capital has three decisions to make: reassess closed-model AI valuations against eroded moat assumptions, reduce unhedged exposure to Chinese open-source derivatives ahead of likely enforcement escalation, and build positions in assets that benefit from the bifurcation, specifically US frontier AI consolidators and GCC infrastructure plays that gain from both the oil windfall and the sovereign AI buildout. The White House has drawn the line. The question is which portfolio positions are on the wrong side of it.