Foreign Bots Pose Threat to Agency Rulemaking, Senator Warns

A top Senate Republican says he wants to prevent AI bots from being used to game the online public comment system at the federal agency in charge of managing US federal lands.

Sen. John Barrasso (R-Wyo.), a frequent critic of public lands policies, is targeting the Bureau of Land Management, which oversees more than 18 million acres in the lawmaker’s state of Wyoming.

His concern: Foreign interest groups can infiltrate the system and AI bots can be used to submit fake comments to BLM. He and three Republican colleagues introduced a bill this summer that would ensure that only U.S. citizens get to submit public comments, and would use a filtering system to prevent AI bots from putting in comments.

The prospects of the bill are uncertain, and it is unclear what steps the agency could take and how prevalent the issue is. Still, the discussion comes amid growing questions about the possibility of artificial intelligence being used to manipulate government and democratic processes.

“We need some thoughtful regulation to promote efficient use of AI and effective use, and also to limit harmful uses, and certainly any kind of malevolent or malicious use of comment forums,” said Nigel Melville, a technology professor at the University of Michigan.

There has been turmoil and controversy with federal agency comments before. In 2017, the Federal Communications Commission’s Restoring Internet Freedom rulemaking process about net neutrality attracted nearly 22 million comments. The office of the New York Attorney General later said that 18 million of those comments were fake and these comments were funded by broadband companies to give the impression of popular support.

The rapid growth of generative AI tools could make it worse, and it’s going to get much more difficult over time to tell the difference between a machine-made comment and a comment from a real human, said Melville.

Wild horses are held in a Bureau of Land Management holding facility after being rounded up in 2017.

Photographer: George Frey/Getty Images

Feedback Matters

Under the Administrative Procedure Act, agencies are required to solicit and consider public comment when developing rules and regulations.

“They have extensive internal procedures for assessing the validity of what information is provided by comments,” said Bridget Dooling, a law professor at Ohio State who specializes in administrative law and regulatory policy.

“It makes a lot of sense for agency staff to be aware of the possibility of mal-attributed or fake comments, whether they’re generated by AI or generated by somebody with a lot of time on their hands,” Dooling said.

In a 2023 paper for the Brookings Institution, Dooling co-wrote that generative AI combined with a mass comment campaign could slow the rulemaking process even more.

“Generative AI disturbs this equilibrium because it may help bad actors generate comments that look more persuasive, i.e., comments that seemingly present evidence beyond mere sentiment,” the paper said.

While the issue of comments created by AI bots crowding out those of constituents could ostensibly affect any federal agency, Barrasso’s bill would specify only that comments under the Federal Land Policy and Management Act come from “citizens of the United States.”

Barrasso is concerned that comments from a bot or a foreign country would be given the same level of considerations as those directly impacted by regulatory changes.

A representative of the Bureau of Land Management, which is an agency within the Interior Department, declined comment when asked about Barrasso’s bill.

But when questioned by Barrasso at a Senate committee hearing earlier this year, Tracy Stone-Manning, the director of the BLM, said: “We read the comments very thoroughly. I think that we can tell when it’s a bot and when it’s a Russian and when it’s an actual Idahoan.”

Filtering Submissions

A 2019 report by the Senate Permanent Subcommittee on Investigations found that most federal agencies don’t have a system in place to address allegations of comments submitted under false identities. The report said at the time that agencies didn’t have consistent policies when it came to screening comments.

But Dooling, the Ohio State professor who previously worked at the Office of Management and Budget’s Office of Information and Regulatory Affairs, pointed out in her 2023 paper that regulations.gov, which is used by most agencies to submit comments, “already implements several techniques to manage large volumes of comments.”

It would be logistically complex to make sure only US citizens were allowed to comment, Dooling said.

“It will almost certainly deter US citizens, as well as others, from commenting,” Dooling said, referring to Barrasso’s bill. “And we have spent decades trying to encourage the public to comment on proposed regulations and things like this that would raise the barriers to commenting should be viewed skeptically and carefully.”