"AI is advancing at a breakneck pace, and every month we keep China and other adversaries from stealing our progress is a month we gain to build the safeguards this technology needs," Rep. George Whitesides said in a statement to CyberScoop.
The AI Cyber Defense Act: NIST tasked to define and deter distillation attacks
A new bipartisan House bill, the AI Cyber Defense Act, would direct the National Institute for Standards and Technology (NIST) to develop voluntary industry standards and best practices aimed at preventing distillation attacks. The legislation, shared with CyberScoop ahead of release, is sponsored by Reps. Pat Harrigan, R-N.C., and George Whitesides, D-Calif. It instructs NIST both to define the term "distillation attack" and to produce guidance that distinguishes those attacks from legitimate sharing and collaboration around open-weight and open-source models.
What the bill counts as an "entity of concern"
The measure identifies four countries whose persons and organizations are presumptively covered: Russia, China, Iran and North Korea. The bill defines "entities of concern" as any foreign person or organization headquartered in one of those four countries, or one operating under the direction of a company based in those countries. The director of NIST would also have the authority to designate specific individuals or organizations as entities of concern.
Audit-ready is a season. It shouldn't be.
Evidence in spreadsheets, controls drifting between audits, frameworks multiplying on flat headcount. Nubivance runs continuous compliance on Rapid7 Cyber GRC - SOC 2, HIPAA, ISO 27001, PCI, CMMC.
End the scrambleHow distillation attacks are described and why sponsors say they matter
The source material defines distillation attacks as cases "when one AI model is trained on the output data and reasoning of another, usually more powerful or advanced, model." Rep. Pat Harrigan told CyberScoop that it is "incredibly expensive" for the U.S. AI industry to maintain a six-to-12 month lead over other countries, and that distillation attacks create an incentive for adversaries to copy models while keeping their own spending low. Harrigan warned that some countries may “hang six to 12 months behind the United States, let us burn really hot and for 5 to 10% of the spend, they can just distill and steal,” enabling a later ramp-up without having borne the same development costs.
Allegations of systematic distillation campaigns and how they operate
The U.S. government, 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. commercial models. According to AI and cybersecurity experts cited in the reporting, these campaigns have involved the use of hundreds of burner and resold accounts and false identities to bypass geographic restrictions to access U.S. frontier models, and have used thousands, sometimes millions, of prompts intended to extract valuable outputs and capabilities for training foreign models.
AI Cybersecurity Roundtable: a NIST forum for defensive practices
The bill would also establish an "AI Cybersecurity Roundtable" at NIST to convene governments and AI industry stakeholders to develop voluntary standards for using AI defensively in cybersecurity for critical infrastructure. Harrigan said he envisions the roundtable including frontier AI companies and software and hardware security experts as independent third-party voices to help identify "an off-the-shelf set of standards" that represents best practices given current technology and threats.
What this means for NIST, frontier AI companies, and creators
- NIST: The agency would be responsible for defining distillation attacks, separating malicious use from legitimate collaboration, producing voluntary standards, and designating specific entities of concern where appropriate.
- Frontier AI companies and security teams (Anthropic, OpenAI, Google and others): They are positioned as both complainants — the reporting notes accusations of systematic distillation campaigns — and potential contributors to the NIST roundtable to help shape defensive best practices and standards.
- Media organizations, creators and artists: The reporting notes that U.S. AI companies themselves face ongoing lawsuits from media organizations, creators, artists and other laborers who allege commercial models were originally trained on their works without compensation or permission — a separate legal pressure point alongside the bill's technical and national-security focus.
The bill centers federal technical guidance on a set of practices and actors defined by nationality and direction, while asking NIST to perform the difficult line-drawing between harmful distillation and beneficial model-sharing. The legislation couples a national-security framing — singling out entities tied to Russia, China, Iran and North Korea — with an explicit call for voluntary, industry-crafted defensive standards and a formal convening mechanism at NIST. Whether that approach will satisfy companies seeking technical clarity, critics focused on legal claims over training data, or lawmakers concerned about global competitive dynamics remains an open question the bill places squarely on NIST's docket.
