New rules spotlight the public sector’s view of AI

Recent US statements underscore the balance governments must strike when it comes to generative AI.

In the news

Just like businesses, the private sector is working to understand how and where generative artificial intelligence fits in—or doesn’t. The big-picture challenges are similar: How do you leverage gen AI’s strengths while minimizing potential damage caused by its flaws? But for government, the tightrope is higher, thinner and without a net.

AI holds enormous promise to boost the productivity of the public sector (and of private providers of public services). But its potential risks—including unintended bias, incorrect output and citizens’ unequal ability to use the technology—abound.

Governments of leading nations must avoid slowing innovation with heavy-handed regulation. In the US, for instance, the Office of Management and Budget (OMB) recently issued new rules that prioritize the public interest and “strike a balance between managing the risks of AI and also encouraging innovation.” The policies offer both specific directives (i.e., all federal agencies must appoint a chief AI officer) and directional guideposts on guardrails.

Another concern, recently echoed in an advisory report to President Joe Biden, is to empower, rather than replace, human scientists.

All of this makes clear governments face the same challenges private enterprises do—with a few other concerns thrown in.

The Cognizant take

The first thing about the OMB announcement that captured the eye of Niels Torm, Cognizant’s GGM Lead Solutions Architect for Privacy, was that it’s well-paired with Biden’s October 2023 executive order on AI and the administration’s 2022 “AI Bill of Rights.” The executive branch is truly trying to put its stamp on AI—and for good reason, he believes.

“AI has powers we’ve never seen before,” Torm says. While the law of unintended consequences is always a factor in both technology advances and regulation, generative AI’s hallucinations—its ability to deliver faulty information that could have drastic consequences for citizens—demand significant government guardrails.

“Monitoring AI in the public sector is just so important,” he says. “If a private company develops an untrustworthy AI, any harm it causes will be contained, and most of the damage will likely be to the company’s reputation.” But if you get a faulty result out of an LLM developed by public authorities, there’s much higher potential for harm. “And not just harm to citizens, but also to trust in public authorities and institutions,” Torm says. As a result, public-sector AI need to be “at a completely different level.”

Moreover, while businesses are free to follow the profit motive, governments are compelled to aim the technology at the common good. In announcing the OMB guidelines, for instance, Vice President Kamala Harris said it’s important to foster innovation to address the climate crisis, he notes. “This a no-brainer. Governments should also use AI to improve healthcare—to drive care to the right people at the right time. Instead of providing healthcare with a shotgun, let’s use a sniper rifle.”

Overall, Torm believes the US is playing an active role in directing and monitoring AI in the public sector, with the hope that the private sector will eventually accept similar oversight. Public opinion will have a big say in that, he says. “If people turn away from companies that use AI in an unethical way, that’ll be regulation of a sort.”

Torm also appreciates the OMB announcement’s emphasis on maintaining human oversight of AI. “Nobody really understands the potential and the risks,” he says. “That will come, but we’re not there yet.”