In a bid to enhance US AI security, the US is reportedly planning to implement measures around advanced AI models, the underlying technology that powers artificial intelligence (AI) systems like ChatGPT, Google Gemini and others.
Citing three people familiar with the matter, a report by news agency Reuters said the US Commerce Department is considering regulations to “restrict the export of proprietary or closed source AI models, whose software and the data it is trained on are kept under wraps.”
These actions align with previous efforts by the Biden Administration to prevent the export of sophisticated AI chips to China. Currently, tech giants like Microsoft-backed OpenAI, Google DeepMind and Anthropic, among others, that have developed AI models can sell them to anyone without government oversight.
Why US may want to restrict export of AI software to China
Researchers have expressed concern that adversaries may misuse these models, which analyse vast amounts of text and images to generate content, to potentially enable cyber attacks or biological weapons.
The Department of Homeland Security said cyber actors are likely to use AI to "develop new tools" to "enable larger-scale, faster, efficient, and more evasive cyber attacks" in its 2024 homeland threat assessment.
Any new export rules could also target other countries, one of the sources said.
The report said that to establish export controls, the US may use a threshold based on the computing power required for model training. Developers reaching this threshold may have to report their plans to the Commerce Department.
Notably, no models have likely hit this threshold yet which may mean that restrictions may be put on unreleased models. Google’s Gemini Ultra is close, according to EpochAI, a research institute tracking AI trends.
Although no final rule proposal exists, considering such measures reflects the US government’s efforts to counter Beijing’s AI plans.
Citing three people familiar with the matter, a report by news agency Reuters said the US Commerce Department is considering regulations to “restrict the export of proprietary or closed source AI models, whose software and the data it is trained on are kept under wraps.”
These actions align with previous efforts by the Biden Administration to prevent the export of sophisticated AI chips to China. Currently, tech giants like Microsoft-backed OpenAI, Google DeepMind and Anthropic, among others, that have developed AI models can sell them to anyone without government oversight.
Why US may want to restrict export of AI software to China
Researchers have expressed concern that adversaries may misuse these models, which analyse vast amounts of text and images to generate content, to potentially enable cyber attacks or biological weapons.
The Department of Homeland Security said cyber actors are likely to use AI to "develop new tools" to "enable larger-scale, faster, efficient, and more evasive cyber attacks" in its 2024 homeland threat assessment.
Any new export rules could also target other countries, one of the sources said.
The report said that to establish export controls, the US may use a threshold based on the computing power required for model training. Developers reaching this threshold may have to report their plans to the Commerce Department.
Notably, no models have likely hit this threshold yet which may mean that restrictions may be put on unreleased models. Google’s Gemini Ultra is close, according to EpochAI, a research institute tracking AI trends.
Although no final rule proposal exists, considering such measures reflects the US government’s efforts to counter Beijing’s AI plans.
