Retirement Plan AI Spurs Evolving Risks for Employer Sponsors

The growing use of artificial intelligence tools in retirement investing is creating new pitfalls for employers that sponsor plans and the fiduciaries that manage those assets, requiring them to meet their legal duties while navigating the changing technological landscape.

AI is increasingly being integrated into retirement benefits operations, including record-keeping, investment advice, and communication with workers who participate in the plans.

That puts an added level of scrutiny on plan sponsors and fiduciaries overseeing investments, as they could trigger Employee Retirement Income Security Act lawsuits if they’re not careful, benefits attorneys said.

“AI is clearly increasing the expectation and the ability of plan sponsors and fiduciaries to test the advice they’re getting” from the tool or outside financial advisers using it, said Stephen Rosenberg, head of the Wagner Law Group’s ERISA litigation practice.

Emerging gray areas stemming from the contrast between the nascent technology and a decades-old law like ERISA include transparency about use of AI tools and liability if information from the technology is inaccurate, attorneys said.

“When something goes wrong, who’s on the hook and who owns it? Are there biases? Are there other challenges?” said David Levine, a principal at Groom Law Group and founder of the AI tool PlanPort.

Duty of Prudence

Marcia Wagner, founder of The Wagner Law Group, said AI can help generate legal documents that the average plan participant can understand, and potentially provide more precise investment advice.

Financial advisers to plans are also using AI tools to research investment costs, benchmark plan performance, and access information that has traditionally been difficult to retrieve, such as details about investment fees or plan performance comparisons.

ERISA’s duty of prudence requires that fiduciaries who oversee plan assets must act with the care and diligence that a knowledgeable person in the same situation would use to protect the interests of plan participants. Using AI tools may violate that standard since the logic they use can be unclear, attorneys said, and fiduciaries using them without proper knowledge could also introduce legal risk.

“There are a number of reasons why the extent to which it will make a difference with respect to a fiduciary’s duty of prudence is not easily determinable at this time,” Wagner said.

But the guiding legal principles of prudence will be the same regardless of whether AI is used, Wagner said. Acting prudently is also context specific, she said, and requires judgment from a human fiduciary.

Rosenberg said plan managers will have to vigilantly monitor for AI hallucinations that provide inaccurate information.

He posed a hypothetical situation where a vendor provides AI-produced information to a fiduciary about expenses or comparing one target date fund to another.

“They’ve got to figure out how they’re going to make sure that it’s right,” Rosenberg said, noting that small errors could spur lawsuits.

Working With Vendors

Melissa Ostrower, co-leader of the employee benefits practice group at Jackson Lewis, said fiduciaries will need to be increasingly on alert about agreements and contracts with vendors and service providers that conduct record-keeping or provide advice. Service provider contracts should be updated to include any uses of AI, she said.

Fiduciaries may have to “ask direct questions and get answers” from vendors about whether they are using AI, how they’re using it, and who is responsible if the tool makes an error, Ostrower said. They must also consider how participant data is used when it’s put into an AI system and evaluate privacy concerns.

Under ERISA, fiduciary decisions are judged by what a knowledgeable expert in the same circumstances would do, Ostrower said. Given how mainstream the technology has become, a prudent expert “can neither ignore AI or use it blindly.”

“Prudent today may not be prudent even in six months or a year because the technology is moving so quickly,” Ostrower said.

Michael Abbott, a partner at Foley & Lardner, said the influx of AI tools will make it even more critical for fiduciaries to document their decision-making process to guard against legal risks.

“Using an AI tool, I don’t care how good it supposedly is, but just using that is not the end of the game for engaging in a proper process,” Abbott said.

Future Use

An August survey for Transamerica’s Prescience 2030 report on employee benefits found 56% agreed or strongly agreed that AI will be integral to plan sponsors’ investment array decisions, while 66% percent agreed or strongly agreed most employers would use AI to identify potential ERISA litigation risks.

Tim Rouse, executive director of the SPARK Institute, a trade group representing the retirement plan industry, said the organization hosted a workshop in March focused on addressing AI in the industry.

“Plan sponsors have been increasingly asking questions as they should be about how AI is being used and under what circumstances,” Rouse said.

The organization established a new AI governance committee following the meeting to address questions members have about how to use the new technology responsibly, he said. Companies will likely use AI to improve efficiency and cut costs, Rouse said, but plan sponsors have a responsibility to be asking questions about how the technology works when it comes to interactions with participants.

Rouse said members will also be watching for any guardrails federal agencies deem appropriate to govern the use of AI in the retirement space moving forward.

A DOL spokesperson didn’t respond to a request for comment about whether the department is considering action.

“Our members are being very cautious and thoughtful about implementing AI tools, and the general public is adopting them very quickly,” Rouse said. “So we’re trying to balance the cautionary application within the industry versus the urgent application by clients and participants.”