Industry calls for US leadership in AI as a democratic imperative

Leading technology companies — including OpenAI, Anthropic, and Amazon Web Services — are hoping that a future artificial intelligence exports program can play a role in leveraging the U.S. AI stack to promote democratic usage of the technology.

Per a July executive order, the Commerce Department's International Trade Administration published a request for information in October calling for industry perspectives on the best format for the American AI Exports Program. After issuing an extension on the comment period, Commerce closed comments on the request Dec. 13.

“Democratic AI is built and distributed to benefit everyone, which is the core value of democracy; autocratic AI will be used by the state to assert control over people,” Benjamin Schwartz, OpenAI’s head of infrastructure partnerships and policy, wrote in his submission. “The world is converging onto two AI stacks. One is championed by the People’s Republic of China (PRC), state-directed, closed, and surveillance-heavy; ours is democratic, market-driven, and safety-aligned. Every country will end up on one of these two stacks, whether they choose to or not. The strategic imperative for the US is to ensure that the democratic stack prevails.”

Schwartz outlined OpenAI’s export policy priorities, which aligned into three major pillars: partnering with the government to advance democratic standards; incentivizing democratic AI and strengthening critical supply chains; and using a consortium model to build a U.S.-led global AI stack.

The proposed consortium would offer international partners a stake in the U.S. tech stack, something OpenAI suggested the ITA Advocacy Center would oversee. OpenAI said that this type of collaboration develops mutually beneficial international partnerships.

“We recommend that the US government drive sustained US leadership in AI by building US-anchored consortia around the world, giving our international partners a stake in the American tech stack, and giving American companies a stake in the success of foreign national champions,” OpenAI wrote.

Anthropic echoed the need for the U.S. tech stack to compete globally, but also stuck to a posture of ensuring its products are used by trustworthy allies. 

“Anthropic strongly supports efforts to advance global adoption of the full American AI stack — including American models and applications, running on American chips, in data centers operated by trusted U.S. companies,” Anthropic’s submission said. 

Anthropic requested the AI exports program work to build trust in U.S. companies, partially through harmonizing standards that ensure an AI product is operating transparently and successfully. It adds that, in the formation of a consortium, Commerce should employ guardrails to ensure U.S. national security isn’t compromised by members, including by making use of a “trusted” partner status. 

Anthropic also highlighted a “key barrier” to U.S. companies participating in the global market: the lack of clarity surrounding text and data mining for AI development. As a solution, the company recommend the new program prioritize engaging with partner governments in the development of a shared framework to recognize non-consumptive computational analysis.

“Fragmented and restrictive copyright regimes in some jurisdictions create significant uncertainty for U.S. exporters and advantage competitors,” Anthropic wrote. “Some partner countries have even considered legal reform that would result in their copyright law applying to U.S. companies even for activity in the U.S.” 

OpenAI echoed the need for regulatory alignment on copyright.

AWS submitted comments agreeing with Anthropic’s push for a trusted partner designation within a consortium and advocated for U.S. cloud providers to take a lead role in the effort.

“AWS stands ready to support the U.S. Government in its effort to advance the American AI Exports Program,” AWS wrote. “U.S. cloud providers are well-positioned to serve as lead entities under the Program by offering a global presence and trusted, full-stack AI solutions with proven resilience and security.”

AWS also expressed its stance against an “overly prescriptive” regulatory regime, advocating instead for “innovation-friendly” AI regulatory regimes abroad, notably through the Economic Diplomacy Action Group within the State Department and Foreign Commercial Service.

On the industry advocacy side, comments from the Business Software Alliance recommended the program adopt a broad perspective on the AI tech stack to include diverse entities in the consortium, addressing regulatory discrepancies that can impede critical data transfer, localization and security concerns. 

“At its core, the AI exports program aims to promote AI products and services, and ITA should ensure that the implementation of the program enhances, not hampers, a competitive AI marketplace,” BSA wrote.