The Department of Commerce, Bureau of Industry and Security (BIS) just announced two developments impacting the Artificial Intelligence (AI) and advanced technologies sectors. Specifically, BIS imposed the first export controls designed to capture leading-edge quantum technology. The new controls focus on computers and related equipment designed to perform calculations at speeds that typical computers are incapable of, as well as technologies used to manufacture advanced semiconductors. BIS also published a proposed rule that would, for the first time, impose requirements and reporting obligations on companies that are developing cutting-edge AI technologies. The AI technologies covered by the proposed rule are large-scale AI models that could easily be modified to perform tasks that pose a risk to national security, defined in the rule as dual-use foundation AI models.
On September 5, 2024, the Department of Commerce announced an interim final rule (IFR) implementing export controls on quantum computing technology and other advanced technologies. The affected technologies are:
The IFR adds Export Control Classification Numbers (ECCNs) to the Commerce Control List to cover the identified advanced technologies and establishes a worldwide license requirement for their export, reexport, or transfer (in-country). The 18 new ECCNS are: 2B910, 2D910, 2E903, 2E910, 3A901, 3A904, 3B903, 3B904, 3C907, 3C908, 3C909, 3D901, 3D907, 3E901, 3E905, 4A906, 4D906, and 4E906.
The IFR is part of the United States’ broader effort to control the export of sensitive technologies. BIS hopes these controls will bring the United States into alignment with allies that have similarly restricted the export of these technologies, such as Japan and the Netherlands. To that end, the IFR introduces a new license exception, License Exception Implemented Export Controls (IEC), which provides that countries that implement similarly stringent export controls on the covered technologies will be exempted from the new worldwide license requirement.
On September 9, 2024, BIS released a Notice of Proposed Rulemaking outlining a new reporting requirement for AI developers and certain computer service providers related to large dual-use foundation AI models and activities exceeding specified, very high computational capability. The proposed rule would obligate such organizations to provide BIS with quarterly reports. Under the proposed rule:
The proposed rule defines dual-use foundation AI models as AI trained on broad data; AI that generally uses self-supervision; AI that contains at least tens of billions of parameters; AI that is applicable across a wide range of contexts; and AI that exhibits, or could be easily modified to exhibit, high levels of performance at tasks that pose a serious risk to security, national economic security, or national public health or safety.
Applicable activities involving these AI models that would require reporting are limited to those involving very high computing capabilities: (i) conducting AI model training using more than 10^26 computational (integer/floating point) operations; or (ii) acquiring or developing a computing cluster with machines transitively connected by data center networking of greater than 300 Gbit/s and having a theoretical maximum greater than 10^20 computational (integer/floating point) operations for AI training.
In announcing the proposed rule, the Department of Commerce stressed that it needs the most accurate, up-to-date information when making policy decisions about the international competitiveness of the industrial base and its ability to support the national defense. The proposed rule follows a pilot survey program conducted by BIS earlier this year, BIS’s Industrial Base Survey, and would amend that process to align with the new quarterly reporting requirements. BIS is soliciting comments on the proposed rule, due October 22, 2024.