Over the course of nine sessions between September and December, a U.S. Senate committee hosted 165 guests including tech industry leaders, academic researchers, venture capitalists, and representatives from academia, labor unions, government agencies, trade associations, and civil society organizations to ask questions and better understand how the government should act on AI. The session topics consisted of innovation; “high impact AI” (such as the use of AI in the health and financial sectors); impacts on the workforce; privacy, liability, and open source; democracy and elections; transparency, explainability, intellectual property, and copyright; “doomsday scenarios”; and lastly, national security. If you want to dive deeper into any specific session, Tech Policy Press has been tracking them all in detail. 

“Over the course of our nine forums, we have learned a great deal about AI, both its potential and its risks,” said Senate Majority Leader Chuck Schumer (D-N.Y.), one of the hosts of the AI Insight Forum, in his closing statements of the last session on Dec. 6, adding that the committee will have next steps to share soon. 

“Now, the work continues here in Congress to draft bipartisan, comprehensive legislation to tackle AI,” he went on. “No question about it: This is exceedingly ambitious. AI is unlike anything we’ve dealt with before, and it may be extremely difficult for legislation to tackle every single issue. Even if we find some solutions or create a degree of consensus to manage AI’s challenges, we must pursue it together.”

Perhaps the biggest takeaway from these sessions is how invested, prepared, and in agreement the senators were throughout. Ahead of the seventh hearing, Sen. Todd Young (R-Ind.) said at Axios’s AI+ Summit that he’s been encouraged and “frankly a little surprised” by the series.

“I expected more disagreement when it came to the role that government should play, the extent to which we should protect workers in various ways from the [AI] technology. There haven’t been very significant disagreements,” he said. 

Attendees feel similarly. Mike Capps, cofounder and CEO of Howso, who attended the seventh session to speak on the issue of explainability, told Eye on AI he didn’t see any contention between the senators, but rather just between the panelists. He was also pleasantly surprised by the senators’ diligence and knowledge around the issues. 

“I was surprised by how strong the questions were and how well-informed the senators were. And I don’t mean that to be rude—it’s just there’s so much going on that they have to do. They were right on point of driving to the issues of explainability,” he said.

It wasn’t long ago when members of Congress were being viciously roasted for asking incredibly misinformed questions during technology-related hearings, such as when Mark Zuckerberg testified before Congress in 2018. In addition to demonstrating a clear understanding of the issues around AI, Capps said he saw around 80 congressional staffers intensely taking notes, as well as many senators who were not part of the proceedings popping in and out just to listen. 

While the decision to keep these sessions closed to the public and media has garnered a lot of criticism, Capps believes this was beneficial for fostering focused and informed discussion. The senators in attendance were the ones who were truly caught up and trying to learn, he said, and no one was trying to get their partisan zinger TV moment—an issue that’s plagued earlier technology hearings, like when Facebook whistleblower Frances Haugen testified before the Senate Commerce Subcommittee on Consumer Protection in 2021, for example. 

From the very beginning, the AI Insight forum was under fire for who was—and wasn’t—being let into the conversation. Attendees invited to the first session on Sept. 13 consisted only of powerful tech executives such as Sam Altman, Mark Zuckerberg, Sundar Pichai, and Elon Musk, among others, whose companies have continuously lobbied against tech regulation and acted recklessly with user data, safety, and their products’ impacts on our society. But what started as a photo opp, as many criticized, expanded to include diverse informed voices. 

“By the time we got to the end,” said Capps, “there were a lot of folks who do real work.”

Eye on AI will be off until the new year. Happy holidays! And until next year, here’s just a bit more AI news. 

Sage Lazzaro
sage.lazzaro@consultant.fortune.com
sagelazzaro.com

A.I. IN THE NEWS

U.K. Supreme Court rules AI can't be named as a patent 'inventor' in landmark case. That’s according to Reuters. The court unanimously denied U.S. inventor Stephen Thaler’s petition to list an AI he created as an inventor on two patents, concluding that “an inventor must be a person” to secure a patent. The U.S. Patent and Trademark Office recently denied Thaler's appeal in a similar case.

FTC bans Rite Aid from using facial recognition for five years after 'reckless' usage. That’s according to Axios. The ban will settle charges that the retailer failed to implement reasonable procedures and prevent harm to consumers in its use of facial recognition technology in hundreds of stores. The agency found the pharmacy’s facial technology repeatedly misidentified individuals, especially women and people of color, as shoplifters. Store employees then acted on the false positive alerts, followed consumers, searched them, publicly accused them, and had them removed by police. Rite Aid’s actions disproportionately impacted people of color, says the FTC.

A top AI image dataset used by leading AI companies contains more than 3,200 instances of CSAM. That’s according to Bloomberg. The Stanford Internet Observatory uncovered the images of suspected illegal child sexual abuse material in the AI database LAION, which has been used to train leading AI image-generators such as Stable Diffusion. Such images have enabled AI systems to transform social media photos of fully clothed real children into nude as well as produce realistic and explicit imagery of fake children. LAION said it’s temporarily removing the dataset.