Earlier this week the U.S. Bureau of Industry and Security restricted the export of artificial intelligence (AI) technologies related to geospatial imagery. Yesterday the Office of Science and Technology Policy told reporters it plans to reveal 10 guidelines regarding the federal government's regulation of AI on January 8 at CES 2020.
Some outlets were already briefed on the guidelines U.S. chief technology officer Michael Kratsios is set to reveal on Wednesday. The White House hasn't publicly released any documents related to the guidelines; we assume they'll be published after Kratsios' presentation tomorrow. But the reporting on these guidelines made it clear that the federal government wants to be careful not to over-regulate AI.
That doesn't come as much of a surprise. This administration has not only been hesitant to introduce new regulations, but it's also rolled back previous rules, including pollution regulations and the net neutrality principle established under the previous administration. It seems unlikely that a burgeoning industry like the one that's sprung up around AI would be treated more strictly.
We've already seen this light touch in action. The restrictions that were revealed on Monday were expected to affect far more products--instead they simply required American companies to obtain licenses before exporting potentially dangerous technologies. (AI related to geospatial imagery could be used for advanced targeting systems, for example, which could pose a national security risk in the wrong hands.)
Reuters said the new guidelines would urge federal agencies to "conduct risk assessment and cost-benefit analyses prior to any regulatory action on AI, with a focus on establishing flexible frameworks rather than one-size-fits-all regulation.” Bloomberg reported that Kratsios also wants any regulations affecting AI to be consistent so companies know what they're allowed to do. He was quoted saying:
“We believe that consistency is really, really important and it sends a very important and very powerful message to industry so that they actually have clarity on the way that they should be thinking about bringing forth products which fall under some sort of regulatory oversight. [...] With this consistency, we think we can spur greater innovation around the country, and we can also help all of our regulators across all of our agencies be more consistent and be more aware of what they need to do to promulgate regulations at a good clip and speed.”
MIT Technology Review's report was the closest anyone came to offering a list of the guidelines themselves, albeit in the form of the reporter's "translation." Those guidelines effectively asked federal agencies to involve the public, industry experts and other agencies when considering regulations; to be flexible, fair and transparent about their approach to AI; and to weigh the societal impacts of AI technologies.