The U.S. Department of Commerce on Monday released a report supporting “open” generative AI models similar to Meta’s Llama 3.1, but advisable that the federal government develop “recent capabilities” to watch such models for potential risks.
The report, written by the U.S. Department of Commerce's National Telecommunications and Information Administration (NTIA), says open weight models expand the supply of generative AI for small businesses, researchers, nonprofits and individual developers. For these reasons, the federal government mustn’t restrict access to open models, the report says – not less than not without examining whether restrictions could harm the market.
The feeling echoes recent comments by FTC Commission Chair Lina Khanwho believes that open models can enable more small players to bring their ideas to market and thus promote healthy competition.
“The openness of the biggest and strongest AI systems will impact competition, innovation and risks of those revolutionary tools,” said Alan Davidson, assistant secretary of commerce for communications and data and NTIA administrator, in an announcement. “The NTIA report recognizes the importance of open AI systems and calls for more lively monitoring of the risks posed by the wide availability of model weights for the biggest AI models. The government has a key role to play in supporting AI development while constructing capability to know and manage recent risks.”
The report comes at a time when domestic and foreign regulators are considering rules that might restrict or impose recent requirements on firms looking for to publish open models.
California is nearby pass Bill SB 1047, which might require any company that has a model with greater than 1026 FLOP of computing power must improve its cybersecurity and develop a strategy to “turn off” copies of the model which can be under its control. Abroad, the EU recently set compliance deadlines for firms under its AI law, which imposes recent rules on copyright, transparency and AI applications.
Meta has said that the EU’s AI policy will prevent them from releasing some open models in the longer term. And quite a lot of startups and huge technology firms have spoken out against the California law that they claim is simply too stressful.
The NTIA's governance philosophy just isn’t entirely laissez-faire.
In its report, the NTIA calls on the federal government to develop an ongoing program to gather evidence on the risks and advantages of open models, evaluate that evidence, and act on those assessments, including imposing certain restrictions on model availability when warranted. Specifically, the report suggests that the federal government examine the security of assorted AI models, support research to mitigate risks, and develop thresholds for “risk-specific” indicators to signal whether a policy change could also be needed.
These and other steps can be consistent with President Joe Biden's executive order on AI, noted Gina Raimondo, U.S. Commerce Secretary. The order called on government agencies and corporations to set recent standards for developing, deploying and using AI.
“The Biden-Harris administration is pulling out all of the stops to maximise the potential of AI while minimizing its risks,” Raimondo said in a press release. “Today's report offers a roadmap for responsible AI innovation and American leadership by advocating openness and making recommendations on how the U.S. government can prepare for and adapt to potential future challenges.”
