# OpenAI’s House Memo — Distillation Leaves the Abuse Queue for U.S.–China Policy

Times of AI Desk · 2026-02-12 · Policy

[https://timesof.ai/2026/02/openai-house-committee-memo-deepseek-distillation](https://timesof.ai/2026/02/openai-house-committee-memo-deepseek-distillation)

> OpenAI (Feb 12) sent the House Select Committee on China a memo—“Updated Stakes for American-Led, Democratic AI”—alleging DeepSeek used multi-stage pipelines and obfuscated third-party routers to harvest U.S. frontier outputs for distillation. Elevates ToS fights into national-security framing; same day as Anthropic’s $30B Series G.

Distillation turns paid API access into a training corpus for rivals. OpenAI’s proprietary escalation: take it **out of private abuse-handling and into a congressional national-security memo**—so DeepSeek’s next models are read as capability transfer, not only clever engineering.

**OpenAI** on February 12 warned U.S. lawmakers that **DeepSeek** and related actors are running **distillation campaigns** against American frontier models. Memo to the **House Select Committee on Strategic Competition between the United States and the Chinese Communist Party**, titled “Updated Stakes for American-Led, Democratic AI” (as cited in Reuters/Bloomberg and later references). Core claim: DeepSeek’s next models should be understood in the context of **distilling** capabilities developed by OpenAI and other U.S. labs. Tactics cited: circumvention of access restrictions; **obfuscated third-party routers** masking query origins; programmatic extraction pipelines. Framing: competitive and national-security issue—capability transfer without equivalent safety investment—not only IP/ToS violation. Context: follows earlier 2025 DeepSeek distillation concerns; positions OpenAI’s anti-distillation hardening as incomplete against adaptive adversaries. Same day Anthropic closed **$30B Series G** at **$380B**—split-screen of U.S. labs raising capital while alleging industrial extraction. Sets up Anthropic’s Feb 23 named-lab report.

## Claims vs checks

Memo existence, title/date, and tactical description are **wire-of-memo** (Reuters/Bloomberg)—full text may not be public. Allegations are **OpenAI’s**; DeepSeek’s rebuttal posture is separate. Do not treat congressional delivery as adjudicated fact.

## Limits

- Secondary reporting on a private memo: quote carefully.
- Obfuscated-router claims are hard for outsiders to reproduce.
- Policy asks in the memo (if any) are not independently enumerated here beyond the distillation framing.

## 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).
- Memo title/date as cited in subsequent congressional and research references: “Updated Stakes for American-Led, Democratic AI,” February 12, 2026.