S-5252-119
Read twice and referred to the Committee on Banking, Housing, and Urban Affairs.
Sponsored by Bill Hagerty (R-TN)
What it does
The BLADE Act would direct the Department of Commerce to assess and publicly list foreign entities ("persons of concern") that conduct "model extraction attacks" — the unauthorized copying of capabilities from closed-source, U.S.-owned artificial intelligence models. It would require the Commerce Department to add identified attackers to the existing export control Entity List, cutting off their access to U.S. technology exports. It would also require the President to impose financial sanctions under the International Emergency Economic Powers Act (IEEPA) on identified persons of concern, blocking their U.S.-based assets and prohibiting transactions with them.
Who benefits
U.S. AI companies that own closed-source models (e.g., large technology firms), whose proprietary model weights and architectures would receive new federal protection. U.S. national security and intelligence agencies, which gain a new threat-assessment mechanism. Domestic AI researchers and startups that compete with foreign AI developers who may have used extracted capabilities. U.S. workers in the AI sector whose employers' competitive advantages would be reinforced. Allied governments and companies that benefit from a stronger U.S. AI competitive position.
Who is hurt
Foreign entities — including researchers, companies, and governments — in designated "countries of concern" (primarily China and Russia) who lose access to U.S. AI models or face sanctions. Foreign persons who are misidentified or listed without adequate due process. U.S. financial institutions and businesses that transact with sanctioned entities and must comply with new restrictions. Academic and research communities that rely on cross-border AI collaboration, which could be chilled by broad definitions. Providers of internet circumvention tools (e.g., VPNs) who may face scrutiny as potential "fraudulent account network providers," despite a narrow carve-out for freedom-of-expression tools.
Supporters argue
Supporters argue that model extraction attacks allow foreign adversaries — particularly China — to shortcut years of costly AI research and development by systematically querying U.S. models to replicate their capabilities, threatening both economic competitiveness and national security. They contend the bill fills a clear enforcement gap: existing export controls cover hardware and software but do not specifically address the theft of AI model capabilities through API abuse. By creating a public "AI Model Extraction Attackers List" and mandating IEEPA sanctions, the bill would impose concrete costs on bad actors and deter future attacks, while explicitly protecting legitimate research conducted within authorized terms of service.
Opponents argue
Opponents argue that the bill's definitions — particularly "model extraction attack" and "person of concern" — are broad enough to sweep in legitimate AI research, competitive benchmarking, and academic study, creating legal uncertainty that could chill beneficial cross-border collaboration. They contend that the mandatory IEEPA sanctions provision removes presidential discretion that has historically allowed for diplomatic flexibility, and that listing entities on the basis of inferred intent (drawn from "volume, structure, pattern, or timing" of queries) without a clear adjudicatory process raises due process concerns. Critics also note that the bill's protections apply only to closed-source U.S. models, potentially entrenching dominant incumbents and disadvantaging open-source AI development.