State AI Task Forces Need to Protect Civil Liberties

Artificial intelligence (AI) as a field of study has been around for decades; however, in the past year, popular new AI products have generated huge consumer, investor, and policymaker interest. The public release of ChatGPT and other generative AI applications has almost instantly created a nation of high-tech tinkerers and future entrepreneurs. Early results show great potential: AI technologies are already helping journalists and researchers to write content, [1] NASA engineers to design technology, [2] scientists to diagnose rare genetic diseases, [3] and insurance companies to price their products. [4]

Government technology vendors are also using—and improving—AI technologies to be adopted by government agencies. These are powerful technologies that, especially when used by the government, require transparency, scrutiny, and training to protect citizens’ civil liberties and maintain their trust in government. Several state legislatures and governors have created AI-focused task forces and advisory bodies to, among other things, inventory state-agency uses of AI technologies and provide recommendations. These are good steps, though state officials should also give advisory bodies actionable and clear areas of focus. The following are some notable areas and recommendations that state AI advisory bodies should consider:

• Anticipate video surveillance everywhere and require warrants
• Prevent mandated installation of video cameras on private property
• Ensure fairness in AI-assisted debanking and financial audits
• Develop policies for the use and sale of commercially available information

Anticipate Video Surveillance Everywhere and Require Warrants

Police departments are using AI systems to analyze hundreds of millions of car trips, via archival license-plate records, to pinpoint and identify likely drug traffickers. [5] Police have also been acquiring video feeds from autonomous-vehicle operators and homeowners to assist with crimefighting: [6] Commercial autonomous vehicles that have multiple sensors and cameras to help drivers are operating in a growing number of cities. Private “doorbell cams” are also common and record video of public roadways. State advisory bodies should consider these various uses and establish training and restrictions on technologies such as car-trip analysis. States should also generally require a warrant or court order before government agencies acquire privately held video footage.

Prevent Mandated Installation of Video Cameras on Private Property

In recent years some cities, including Atlanta and Houston, [7] have required the installation of cameras on private property and in private establishments, including convenience stores and bars. They have also required law-enforcement access to recorded video. Mandated installation of de facto government surveillance systems on and in private property raises many concerns and will tend to decrease public trust. Further, these installation mandates likely violate the US Constitution’s proscription against government taking of private property. In a different context, the Supreme Court held that government-mandated installation of cable TV wires and boxes on private property was an unconstitutional taking in Loretto v. Teleprompter Manhattan CATV Corp. [8] While there’s nothing objectionable with businesses assisting law enforcement, installation of and access to government-mandated surveillance cameras should be denied.

Ensure Fairness in AI-Assisted Debanking and Financial Audits

AI technologies are increasingly used to assess financial risk and ensure tax compliance. In order to avert regulatory scrutiny, some banks are reportedly using AI and social network analysis to determine whether to “debank” someone. [9] It was also reported recently that the IRS is using AI systems to help identify taxpayers for audits. [10] The use of AI systems in banking and in federal– and state–tax compliance could mean better banking services and fairer, more effective tax collection. But it could also mean selective enforcement and escalating political lawfare—the use of the legal systems to delegitimize an adversary. State advisory bodies should consider the use of AI systems in banking and tax audits and recommend policies to promote fairness and guard against potential abuse of these new tools.

Develop Policies for the Use and Sale of Commercially Available Information

Federal and state agencies are increasingly acquiring and using sensitive commercial and public information for law enforcement and general societal surveillance. A recently declassified report to the Director of National Intelligence, for instance, noted increasing government acquisition and use—absent a warrant—of “commercially available information” of citizens. [11] This information includes social security numbers, employment history, geolocation, and social media posts. The report notes that commercially available information “is increasingly powerful for intelligence and increasingly sensitive for individual privacy and civil liberties, and the [intelligence community] therefore needs to develop more refined policies to govern its acquisition and treatment.” The report adds, “Even subject to appropriate controls, [commercially available information] can increase the power of the government’s ability to peer into private lives to levels that may exceed our constitutional traditions or other social expectations.”

The acquisition and use of public (e.g., social media) and commercially available content by state governments and law enforcement should at the very least have approval from electorally accountable officials such as state attorneys general. Also, advisory committees should evaluate current and likely uses of such information and recommend best practices to ensure that uses of novel AI technology do not erode the public’s expectations of privacy.

Notes

1. David Nield, “5 Ways ChatGPT Can Improve, Not Replace, Your Writing,” Wired, July 23, 2023.

2. “How NASA Is Using AI To Design Aerospace Parts,” Aviation Week Network, March 7, 2023.

3. Brent Skorup, “AI in Medicine: Companies and Stories to Watch,” Digital Spirits, March 7, 2023, https://digitalspirits .substack.com/p/ai-in-medicine-companies-and-stories.

4. Jeffrey Sonnenfeld and Steven Tian, “Speculative Headlines about Artificial Intelligence Are Missing the Point. Here’s What America’s Top CEOs Told Us They Intend to Use A.I. For,” Fortune, June 21, 2023.

5. Thomas Brewster, “This AI Watches Millions of Cars Daily and Tells Cops If You’re Driving Like a Criminal,” Forbes, July 17, 2023. Credit to my colleague Ryan Dellinger for flagging this story.

6. “As self-driving cars become a fixture in major American cities like San Francisco, Phoenix, and Los Angeles, police are increasingly relying on their camera recordings to try to solve cases.” Julia Love, “Police Are Requesting Self-Driving Car Footage for Video Evidence,” Bloomberg, June 29, 2023.

7. See Atlanta City Council, “Atlanta City Council Approves Legislation to Require Service Stations to Install Video Surveillance Systems,” news release, August 7, 2023.

8. Loretto v. Teleprompter Manhattan CATV Corp., 458 U.S. 419 (1982).

9. Izabella Kaminksa and Dan Bloom, “Nigel Farage and the Great Anti-Brexit Unbanking Plot,” Politico, July 5, 2023.

10. Alan Rappeport, “I.R.S. Deploys Artificial Intelligence to Catch Tax Evasion,” New York Times, September 8, 2023.

11. Office of the Director of National Intelligence Senior Advisory Group, “Panel on Commercially Available Information,” January 27, 2022.