WASHINGTON — In an aggressive move to safeguard the United States’ technological supremacy, a newly introduced bipartisan House bill aims to task the federal government with developing formal, standardized defenses against artificial intelligence "distillation attacks."

The legislation, titled the AI Cyber Defense Act, was shared exclusively with CyberScoop prior to its official introduction. Sponsored by Representatives Pat Harrigan (R-N.C.) and George Whitesides (D-Calif.), the bill would direct the National Institute of Standards and Technology (NIST) to formulate voluntary industry standards and best practices. The primary objective is to harden American frontier AI models against intellectual property theft and unauthorized model extraction by entities rooted in adversarial nations, specifically naming China, Russia, Iran, and North Korea.

The bill arrives at a precarious geopolitical crossroads. As generative AI shifts from a novelty to a foundational pillar of modern infrastructure, economy, and national security, policymakers are increasingly viewing technological advantage not just as a commercial race, but as a core component of geopolitical defense.

Main Facts

The core of the AI Cyber Defense Act centers on combating a specific form of intellectual property appropriation known as model distillation.

• The Sponsor Coalition: The bill is co-sponsored by Rep. Pat Harrigan (R-N.C.) and Rep. George Whitesides (D-Calif.), demonstrating rare bipartisan consensus on technology policy and national security.
• The Directive to NIST: If passed, the legislation forces NIST to establish voluntary industry guidelines, define what constitutes a distillation attack, and build a protective framework specifically aimed at mitigating risks from "entities of concern."
• Entities of Concern: The bill explicitly defines these entities as organizations or individuals headquartered in, or directed by actors within, four nations: Russia, China, Iran, and North Korea. NIST’s director would also be granted discretionary power to designate specific third-party organizations as entities of concern.
• The AI Cybersecurity Roundtable: The legislation mandates the creation of an ongoing advisory body at NIST, comprising government officials, frontier AI companies, and cybersecurity experts specializing in hardware and software. This roundtable will help shape practical, "off-the-shelf" defensive standards for critical infrastructure.

Chronology and Context: The Rise of the Distillation Threat

While generative AI has advanced at breakneck speed over the last half-decade, the mechanics of how foreign adversaries acquire these capabilities have evolved from traditional espionage to sophisticated, automated data harvesting.

The Evolution of Model Extraction

In the early days of generative AI, geopolitical rivals focused primarily on attempting to train models from scratch, hampered by hardware export controls (such as U.S. restrictions on high-end GPUs). However, as American companies like OpenAI, Anthropic, and Google invested billions of dollars into research, compute power, and massive dataset curation, a cheaper alternative emerged: model distillation.

Recent Escalations

Over the past 12 to 18 months, U.S. intelligence agencies and top AI labs have flagged systematic efforts by foreign commercial entities to clone American capabilities.

• The Accusations: The U.S. government, alongside major commercial labs including Anthropic, OpenAI, and Google, has repeatedly accused Chinese firms—such as Moonshot AI, DeepSeek, and Alibaba—of orchestrating coordinated campaigns to distill U.S. frontier models.
• The Mechanics: According to cybersecurity researchers, these operations typically leverage hundreds of burner accounts, proxy networks, and synthetic identities to bypass geographic restrictions. Once inside the ecosystem, these actors fire thousands or millions of targeted prompts designed to extract the internal reasoning, weights, and fine-tuned capabilities of advanced models, funneling that output back into domestic foreign projects at a fraction of the original cost.

Legislative Timing

The drafting of the AI Cyber Defense Act follows a mounting series of federal warnings regarding intellectual property erosion. By formally tasking NIST with tackling this issue, Congress is attempting to translate ad-hoc cybersecurity warnings into a cohesive, national defensive strategy.

Supporting Data and Economic Realities

The economic disparity at the heart of the legislation is stark. Training a frontier foundational model requires hundreds of millions of dollars in capital expenditure, thousands of specialized semiconductor chips, and years of proprietary research.

According to Rep. Harrigan, maintaining the United States’ fragile six-to-12-month lead over foreign competitors is "incredibly expensive."

[U.S. Frontier R&D Investment] ---> Burns heavy capital & research hours
                                      │
                                      ▼
                      [Adversarial Distillation] 
                      (Costs 5-10% of total spend)
                                      │
                                      ▼
                      [Short-term Free Riding] ---> [Potential Market Leapfrog]

The "Free-Rider" Strategy

Harrigan argues that adversarial nations—particularly China—are making a deliberate, calculated economic choice. Rather than burning capital on foundational research, these nations are content to trail American companies by a few months, siphoning off innovations via distillation attacks while spending only a fraction (an estimated 5% to 10%) of the total cost.

"Where they actually just see it as more beneficial for themselves in the short term, to 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," Harrigan explained in an interview with CyberScoop.

The ultimate fear among national security hawks is that this dynamic creates a trap: American companies will exhaust their financial and research capital keeping the industry engine burning, only for foreign competitors to leapfrog them once market fatigue sets in.

Domestic Complications

Compounding the challenge, the U.S. AI industry is simultaneously fighting battles on the home front. Commercial developers are currently targets in multiple high-profile copyright lawsuits filed by media organizations, artists, authors, and labor unions, who argue that foundational models were built using their intellectual property without consent or compensation. This puts U.S. labs in a defensive posture from both legal and foreign-espionage angles.

Official Responses and Stakeholder Perspectives

The introduction of the bill has drawn praise from lawmakers and technical experts alike, who view standard-setting as an essential first step toward industrial resilience.

• Rep. George Whitesides (D-Calif.): Emphasizing the urgency of the moment, Whitesides noted in a statement, "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. Pat Harrigan (R-N.C.): Highlighting the strategic threat, Harrigan underscored that distillation attacks fundamentally distort market incentives by rewarding theft over true innovation. He envisions the proposed NIST "AI Cybersecurity Roundtable" as an actionable resource where critical infrastructure operators can find reliable, peer-reviewed defense strategies. "A place where you can go and pick an off-the-shelf set of standards that says these are the best practices as we know them all today," Harrigan said.
• The Technical Challenge: One of the most delicate tasks assigned to NIST under the bill will be definitional. The agency must carefully separate malicious distillation attacks from the legitimate, collaborative data-sharing ecosystems of open-weight and open-source models. Open-source innovation has been a vital engine for general AI research, and over-regulation risks stifling domestic progress while attempting to stop foreign theft.

Broader Implications for Critical Infrastructure and National Security

The ramifications of the AI Cyber Defense Act extend far beyond the balance sheets of Silicon Valley tech giants. As artificial intelligence becomes deeply integrated into energy grids, financial systems, healthcare networks, and military logistics, securing the underlying models is tantamount to securing national sovereignty.

The Role of NIST Standards

While the standards produced by NIST under this act would initially be voluntary, federal guidelines often serve as the de facto baseline for federal procurement and private-sector compliance. By creating a formalized framework for spotting and blocking distillation attempts, the legislation aims to give domestic enterprises the tools they need to lock down their API endpoints, monitor unusual prompt behaviors, and identify synthetic traffic patterns.

Future Outlook

As Congress moves forward with the AI Cyber Defense Act, the technology sector will be watching closely to see how NIST navigates the fine line between open scientific collaboration and aggressive intellectual property protection. If successful, the legislation could provide the structural armor needed to protect American innovation from becoming a casualty of its own success.