Skip to main content
Live
Main content

House report urges removal of Chinese open-weight AI models as Cursor weighs exit

A 12-recommendation joint report says Anysphere plans to explore phasing out PRC models, and that Airbnb and DoorDash rely on them too.

Jaeden Schafer
Editor in Chief · · 5 min read
House report urges removal of Chinese open-weight AI models as Cursor weighs exit

The House Committee on Homeland Security and the House Select Committee on China released a joint report on November 20 urging the removal of Chinese-developed open-weight AI models from federal systems and high-risk critical infrastructure. The report also disclosed that Anysphere, the company behind the Cursor coding platform, told the committees it plans to explore phasing out PRC-developed open-weight models because of geopolitical risk.

Chairmen Andrew Garbarino (R-NY) and John Moolenaar (R-MI) opened the investigation on April 29, 2026. The report frames Anysphere's position as a plan to explore a phase-out, not a confirmed decision, and no replacement timeline has been published. Cursor is owned by SpaceX, which acquired Anysphere in August.

The Cursor question centers on Composer 2, the flagship proprietary model in Cursor. According to the report, Anysphere built Composer 2 on a base model from Moonshot AI, the developer of Kimi, and that fact became public only after an independent developer identified the base model.

Key facts

  • 01Two House committees urged removing PRC-developed open-weight models from federal systems and high-risk infrastructure.
  • 02Cursor maker Anysphere told the committees it plans to explore phasing out PRC open-weight models.
  • 03Cursor's Composer 2 model was built on a Moonshot AI base model, the report says.
  • 04PRC models made up 41% of downloads on the leading open-model repository over the prior year.
  • 05The report makes 12 recommendations, starting with a national American open-weight AI initiative.

Anysphere is not the only US company named. The report says Airbnb uses Alibaba's Qwen models in customer support, including an assistant that handles roughly 80,000 guest and host interactions a day. DoorDash told the committees that 25 PRC model families had been authorized and deployed in a non-trivial way, and that replacing one embedded PRC model could take up to six months while most other uses could switch almost immediately.

The committees found that PRC models accounted for 41% of downloads on the leading open-model repository over the prior year, and that the highest-ranked US open-weight model placed 10th on the Artificial Analysis index, behind six PRC models.

Price is a large part of the pull. The report's pricing table puts Qwen3-235B-A22B-Instruct at about $1.15, DeepSeek-R1 at about $2.74 and Kimi K2.6 at about $4.95 for a million input and a million output tokens, against about $30 for Claude Opus 4.8 and about $35 for GPT-5.5. The report notes those figures are not adjusted for capability, caching, batch discounts or self-hosting.

The report makes 12 recommendations. They include a national American open-weight AI initiative, a model provenance standard and AI bill of materials, extending federal restrictions beyond DeepSeek to bar PRC-developed models from federal systems and contract work, and making CISA the single point of submission for AI-related cyber incident reports.

Other recommendations target illicit distillation through sanctions and Entity List designations, an examination of sales of American training data to PRC labs, and governance of both closed and open models without overregulation. The committees said they will continue their oversight of these issues, and a classified annex is available to cleared personnel.

Related · from this week
OpenAI safety veteran David Robinson resigns, says culture is 'broken'
Jaeden Schafer · 4 min read →
ShareXLinkedInEmail
AI Box

Every AI model. One chat.

The latest models from ChatGPT, Claude, Gemini, Sora, ElevenLabs — 80+ models in a single chat. Compare answers side by side. Pick the best one every time.

  • ChatGPT, Claude, Gemini, Grok, DeepSeek — in one chat
  • Generate images & video with Sora, Veo, Ideogram
  • Compare any two models side by side
  • From $8.99/mo · 80+ models, all included
Try AI Boxaibox.ai
Trusted by 3,000+ teams
Got a tip?

Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.

Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.

AI Box Daily briefingFree · Daily · No fluff

Stay ahead of everyone in AI.

The tightly edited AI news email engineers, founders, and investors actually open. One email. Every weekday. Five minutes to finish.

Loved by 10,000+ AI professionals
Free forever. Unsubscribe with one click.

The briefing read inside teams at

Keep reading

More from Security

OpenAI logo
Security

OpenAI safety veteran David Robinson resigns, says culture is 'broken'

Robinson, who led safety reports for major product launches over 3.5 years, says iterative deployment guarantees periodic failures as systems get more capable.

Jaeden Schafer4 min read
Nvidia logo
Security

Nvidia opens OpenShell and Sentry to contain rogue AI agents

The chipmaker is pushing an open-source Agent Safety Platform as frontier labs disclose agents hacking Hugging Face and probing government sites.

Jaeden Schafer4 min read
Anthropic logo
Security

DC Circuit upholds Trump blacklist of Anthropic in 2-1 ruling

Appeals court says Pentagon can designate Anthropic a supply chain risk for restricting Claude features, even absent malicious intent.

Jaeden Schafer5 min read