OpenAI Warns Congress DeepSeek Is Distilling U.S. Frontier Models
On February 12, 2026, OpenAI sent the House Select Committee on Strategic Competition between the United States and the Chinese Communist Party a memo titled “Updated Stakes for American-Led, Democratic AI,” alleging DeepSeek used sophisticated, multi-stage pipelines—including obfuscated third-party routers—to harvest model outputs for distillation. The same-day letter framed industrial-scale free-riding as a national-security and competitive threat ahead of expected Chinese model releases.
TLDR
OpenAI on February 12, 2026 warned U.S. lawmakers that DeepSeek and related actors are running distillation campaigns against American frontier models. In a memo to the House Select Committee on China, OpenAI said it observed accounts tied to DeepSeek employees building methods to circumvent access restrictions, route through obfuscated third-party systems, and programmatically harvest outputs for training competitors—describing ongoing free-riding on U.S. lab capabilities.
What the memo alleged
From contemporaneous Reuters/Bloomberg reporting on the memo (publicly referenced as “Updated Stakes for American-Led, Democratic AI”):
- Audience: House Select Committee on Strategic Competition between the U.S. and the Chinese Communist Party.
- Core claim: DeepSeek’s next models should be understood in the context of distilling capabilities developed by OpenAI and other U.S. frontier labs.
- Tactics cited: Circumvention of access restrictions; obfuscated third-party routers masking query origins; programmatic extraction pipelines for distillation data.
- Framing: Not only IP/ToS violation, but a competitive and national-security issue—capability transfer without equivalent safety investment.
- Context: Follows earlier 2025 concerns about DeepSeek distillation of U.S. models; positions OpenAI’s own anti-distillation hardening as incomplete against adaptive adversaries.
The memo landed the same day Anthropic closed its $30B Series G at $380B—a split-screen of U.S. labs raising capital while alleging industrial extraction of their model outputs.
Why this story matters
Distillation turns paid API access into a training corpus for rivals. OpenAI’s congressional memo elevates the issue from private abuse-handling to U.S.–China AI policy, setting up later February disclosures (Anthropic’s Feb 23 DeepSeek/Moonshot/MiniMax report) and 2026 executive/legislative responses on industrial-scale model theft.
Sources
- Reuters: OpenAI warns lawmakers DeepSeek targeting U.S. models for distillation (February 12, 2026).
- Bloomberg: OpenAI memo to House China committee on DeepSeek distillation tactics (February 12, 2026).
- OpenAI memo title/date as cited in subsequent congressional and research references: “Updated Stakes for American-Led, Democratic AI,” February 12, 2026.
Featured Image Alt Text
OpenAI logo and congressional seal over abstract distillation arrows from U.S. models to a DeepSeek-labeled system.
Tags
OpenAI, DeepSeek, Distillation, Congress, China, National Security, Policy