Bipartisan Bill Would Direct NIST to Develop Guidance on AI Distillation Attacks
Summary
The proposed AI Cyber Defense Act would direct NIST to develop voluntary guidance against distillation attacks and establish a roundtable on using AI to defend critical infrastructure. The article reports allegations that Chinese firms have conducted campaigns to copy U
Key points
- The proposed AI Cyber Defense Act, sponsored by Reps. Pat Harrigan and George Whitesides, would direct NIST to develop voluntary standards and best practices against distillation attacks.
- The bill would require NIST to define distillation attacks and distinguish them from open-weight and open-source model collaboration.
- The article says the U.S. government and AI companies Anthropic, OpenAI and Google have accused Chinese companies including Moonshot AI, DeepSeek and Alibaba of systematic campaigns to distill U.S. commercial models.
- Experts cited in the article say the campaigns use burner and resold accounts, false identities and large volumes of prompts to obtain model outputs and capabilities.
- The bill would establish an AI Cybersecurity Roundtable at NIST to develop voluntary standards for using AI to defend critical infrastructure.
- The bill defines entities of concern by links to Russia, China, Iran or North Korea, and would let the NIST director designate specific individuals or organizations.
Article Details
- Event Type
- Proposed bipartisan legislation to develop federal guidance and voluntary standards for defending AI models against distillation attacks, and to establish an AI cybersecurity roundtable at NIST.
- Impact
- The bill would direct NIST to define distillation attacks and develop guidance to help prevent foreign entities from copying U.S. commercial AI models. It would also convene government and AI-industry stakeholders to develop voluntary standards for defensive AI use in critical-infrastructure cybersecurity. The article reports allegations by the U.S. government and AI companies that Chinese companies have conducted campaigns to distill U.S. models; the bill is proposed, not enacted.
People
George WhitesidesU.S. representative and sponsor of the AI Cyber Defense Act; said the bill would help protect U.S. progress in AI.Pat HarriganU.S. representative and sponsor of the AI Cyber Defense Act; discussed the bill and the concern that distillation attacks could enable copying of U.S. AI models.
Vendors
Anthropicgovernment, along with AI companies Anthropic, OpenAI and Google, has accused Chinese companies like Moonshot AI, DeepSeek and Alibaba, of carrying out “systematic” campaigns to distill and copy U.S.Googlegovernment, along with AI companies Anthropic, OpenAI and Google, has accused Chinese companies like Moonshot AI, DeepSeek and Alibaba, of carrying out “systematic” campaigns to distill and copy U.S.OpenAIgovernment, along with AI companies Anthropic, OpenAI and Google, has accused Chinese companies like Moonshot AI, DeepSeek and Alibaba, of carrying out “systematic” campaigns to distill and copy U.S.
Products
Countries
Chinaand best practices for preventing distillation attacks from foreign entities from four countries: Russia, China, Iran and North Korea.Iranpractices for preventing distillation attacks from foreign entities from four countries: Russia, China, Iran and North Korea.North Koreafor preventing distillation attacks from foreign entities from four countries: Russia, China, Iran and North Korea.Russiaand best practices for preventing distillation attacks from foreign entities from four countries: Russia, China, Iran and North Korea.