Biden Admin Mulls Export Restrictions For AI Models
The United States is considering safeguarding American artificial intelligence (AI) models from access by hostile nations such as China and Russia.
Reuters, citing three people familiar with the matter, reported that the Biden administration is drawing up preliminary plans that would place guardrails around the most advanced AI models, which are the foundation of AI systems such as ChatGPT.
The US is tightening the screw on countries such as China and Russia. Earlier this week the US Department of Commerce informed both Intel and Qualcomm that, effectively immediately, it had revoked some of their export licences to customers in China.
Intel said the US move would impact its second quarter revenue below the midpoint of its forecast.
Export restrictions
Former President Donald Trump’s administration had added Chinese giant Huawei and others to the US Entity List in May 2019.
Essentially, US firms have to apply for permission or a license to sell anything to Huawei, and certain other Chinese firms.
In January 2023 the US Commerce Department reportedly stopped providing export licenses for Huawei.
In October 2023 the Biden Administration tightened export controls on advanced AI semiconductors to China.
AI model restrictions
Now according to the Reuters report, the US Commerce Department is considering a new regulatory push to restrict the export of proprietary or closed source AI models.
This action would complement the previous US measures to block the export of sophisticated tech China in an effort to slow Beijing’s development of the cutting edge technology for military purposes.
The Commerce Department declined to comment while the Russian Embassy in Washington did not immediately respond to a request for comment, Reuters reported.
The Chinese Embassy described the move as a “typical act of economic coercion and unilateral bullying, which China firmly opposes,” adding that it would take “necessary measures” to protect its interests.
Hostile nations
Reuters noted that at the moment, there is nothing to halt the likes of Microsoft-backed OpenAI, Alphabet’s Google DeepMind, and rival Anthropic, from selling powerful AI models to any customer in the world.
The sale of these AI models, without government oversight, raises the risk of these cutting edge AI models falling into the hands of hostile nation states, which could in turn use the tech for disinformation campaigns, to wage cyber attacks, or even create potent biological weapons.
In January the UK’s National Cyber Security Centre (NCSC), part of GCHQ, warned that increasingly pervasive AI systems “will almost certainly increase the volume and impact of cyber attacks in the next two years.”
One of the sources told Reuters any new export control would likely target Russia, China, North Korea and Iran.
Computing threshold
To develop an export control on AI models, the sources said the US may turn to a threshold contained in an AI executive order issued last October by President Biden, that is based on the amount of computing power it takes to train a model.
When that level is reached, a developer must report its AI model development plans and provide test results to the Commerce Department.
That computing power threshold could become the basis for determining what AI models would be subject to export restrictions, Reuters reported, citing two US officials and another source briefed on the discussions.
If used, it would likely only restrict the export of models that have yet to be released, because currently AI models are not thought to have reached the threshold yet.
Reuters reported that Google’s Gemini Ultra is seen as being close, according to EpochAI, a research institute tracking AI trends.
The US Commerce Department is reportedly far from finalising a rule proposal, the sources told Reuters. But the fact that the move is already being considered shows how the United States is already planning to further tighten restrictions on hostile nations.
