10
Loading video...

By Jesus Mesa
Reporter
0
Share
Newsweek is a Trust Project member
See more of our trusted coverage when you search.Prefer Newsweek on Googleto see more of our trusted coverage when you search.
A staffer for House Minority Leader Hakeem Jeffries built an artificial intelligence (AI) program that combed through Democratic lawmakers' posts and drafted lines for his boss to use, according to a report published Thursday.
The program, which the staffer named Caucus Pulse, surfaced one day after Democrats voted 251-11 to nominate Jeffries, of New York, for speaker. The Brooklyn-born Democrat spent a good chunk of the fall telling Congress to regulate AI, and come January, his party will control the House and Senate.
It's an awkward spot for a party that leaned on AI skepticism to win the November 3 midterm elections. Several experts Newsweek talked with saw no real conflict in using a tool while writing rules for it, as long as offices are upfront about how they use it.
Read More on News
"Technology itself is not the problem," Rishi Bharwani, U.S. director of the tech policy group Reset Tech, told Newsweek. "These tools are becoming increasingly essential to how we work, communicate and participate in public life."

...
What the Jeffries Tool Did
The Washington Sun, which first reported on the program, found its source code in a public GitHub repository. According to the Sun, Caucus Pulse was built largely with Anthropic's Claude and paid X to download posts from hundreds of accounts, mostly House lawmakers and journalists. The code shows it went through 114,647 posts over roughly six weeks.
Per the Sun, it kept tabs on caucus factions, flagged racist attacks on Jeffries and logged which members defended him. It even wrote one-liners he could throw at Republicans.
In a September memo, after a former Anthropic researcher resigned with a public warning about AI, Caucus Pulse advised Jeffries to pursue "hearings/oversight" on the technology "rather than a bill." The program also suggested that he avoid echoing the researcher's concerns given the company's donations to Democrats, according to the Sun.
Related Story

Bigger AI Models Aren’t Always Better
Jeffries' office said he did not know about the program and never used its material.
"Just like people from all walks of life do every day, a staffer in this office conducted a short-lived experiment with AI in a personal capacity, without the involvement of the Office of the Democratic Leader," spokesperson Christie Stephenson said. "No official resources were used, and we're addressing this matter internally."
The office told the Sun that the House permits official AI use and that staff may use AI to monitor and collate public information. According to data the office provided, it has spent about $5,300 on OpenAI and Anthropic tools since the current Congress began in January 2025.
Jeffries pushed hard on AI this fall. He told ABC's This Week in September that Congress needed "to begin to act urgently," and later dismissed President Donald Trump's voluntary AI safety pact as "entirely unenforceable."
Newsweek reached out to Jeffries' office via email for additional comment.

...
How Americans Feel About AI
The public has been moving the other way. Pew Research Center found in June that 52 percent of U.S. adults were more concerned than excited about AI in daily life. In 2021, it was 37 percent. Only 9 percent leaned excited.
Democrats were more uneasy still, at 56 percent, up from 46 percent in 2023. About half of Republicans said the same, per Pew.
Gallup has picked it up, too, as 39 percent of Americans now think AI does more harm than good, compared with 31 percent last year.
On Election Day, NBC News asked voters how much AI and data centers weighed on their pick for Congress. Nearly one-quarter said crucial and another one-third said major, and those voters went for Democrats 64 percent to 34 percent. Fifty-seven percent figured AI would do society more harm than good. Three in four didn't want a data center in their community, a fight that had been reshaping key Senate races for months.
Split the Experts Saw Coming
In October 2025, data scientist Nathan E. Sanders and security technologist Bruce Schneier wrote in Time that AI would split the major parties. They expected the gap to widen, "resulting in AI being predominantly used by one political side in the 2026 elections," and said Republicans looked "poised to exploit the technology."
A year later, Sanders, an affiliate of Harvard's Berkman Klein Center, sees no contradiction in Democrats doing both.
"Members of Congress owe it to their constituents to constantly challenge themselves about how to leverage new technologies to serve them and the broader public better, while simultaneously legislating on the regulation of those industries and holding industry leaders accountable," he told Newsweek in an email. "AI is no different from older digital technologies in that way."
This spring, a survey by the American Association of Political Consultants found that 64 percent of Republican consultants used AI daily, compared with 49 percent of Democrats, according to Axios.
"There are many Democrats who are using AI, but there's also a lot of skepticism," Democratic strategist Larry Huynh told the outlet. "I don't see the same level of skepticism on the Republican side."
Republican committees went all-in on AI-made ads. Democrats mostly kept the tools in the back office, for drafts and data, while running against the industry out front. One ad by Texas Democratic state Representative James Talarico hit Republican Attorney General Ken Paxton, his U.S. Senate opponent, for "selling off Texas" to data centers, then closed with the line "Made by humans, not AI."
Data centers hurt candidates in both parties. By Newsweek's count, 16 candidates who backed data center-friendly policies were sitting in toss-up races heading into November. Even so, Republican leaders largely stuck with Trump on the buildout.
"I think the 2026 elections demonstrated that voters are looking for a different kind of politics," Bharwani said, pointing to Talarico in Texas and Democratic Senator Jon Ossoff of Georgia [both of whom won]. "Candidates need to be able to explain how these issues affect people's lives rather than simply repeating talking points about innovation or attacking the technology itself."

...
The Hypocrisy Charge and the Money
Conservative outlets jumped on the Sun's story, and so did progressives who already had doubts about Jeffries. Truthout set his AI warnings next to his tech donors: $150,000 from executives at Anduril, a defense tech company, and $44,300 from a Palantir executive.
Jason Goldman, a former chief digital officer in the Obama White House, took a lighter view, writing on X that without the word "surveillance," the tool was "probably a good idea depending on implementation." He added: "The job of doing clips is a well established one in political comms."
AI money was everywhere this election cycle. Industry-linked super PACs put $79 million into 68 House and Senate seats, according to Tech Influence Watch data cited by Fortune, the industry's biggest year yet.
Super PAC Leading the Future, funded in large part by OpenAI President Greg Brockman and Andreessen Horowitz, spent on candidates who wanted light rules and one national standard, while Public First, with $20 million from Anthropic, backed people pushing for more safeguards.
Leading the Future's biggest general election bet went bust in South Carolina, where groups tied to it spent about $4.8 million backing Republican Senator Darline Graham. Democrat Annie Andrews beat her 51.1 percent to 46.7 percent. Across the map, data centers were a flashpoint in races like that one.
"One of the biggest challenges is that we're discussing regulation at a time when enormous amounts of money are flowing into political campaigns from industries that have a direct stake in these decisions," Bharwani said. "That makes it particularly important for candidates to be clear about where they stand and whose interests they are representing."
Can Lawmakers Use AI Ethically?
Asked whether a party that wants to regulate AI can also use it, Hillary Lehr of Higher Ground Labs, a firm that invests in progressive political technology, said yes.
"In fact it's essential," she told Newsweek. "If elected officials and their teams can use technology responsibly to do their jobs better, they should, and taxpayers should expect them to."
She argued that regulating frontier AI companies and using AI responsibly serve the same goal.
"Responsible use will require transparency and respect for people's data and privacy, not avoidance," Lehr said.
Asked about the Jeffries program, Sanders called it "very reasonable" for a leadership office to track public posts and sentiment with AI. Campaigns have done the same with "social listening" software for years, he said.
"I think it's generally a good thing for political leaders to be well informed of what their colleagues are saying and how the public is responding," he wrote.

...
His line is about data. Sensitive or nonpublic constituent information belongs only on platforms that meet strict security standards, he wrote, and "staffers should not go into personal AI tools that are outside of the governance of the responsible offices" and use them with that kind of data. He pointed to a model AI policy for congressional offices published by the POPVOX Foundation.
Bharwani brought up social media. People know the risks and use it anyway to reach more people, he said.
"There is nothing inherently wrong with campaigns using these tools," he said. "The important question is how they are being used and whether there are appropriate safeguards in place."
Congress has done this before, Lehr argued.
"Lawmakers drive cars and mandated seatbelts and crash testing," she said. "Candidates run TV ads and passed 'stand by your ad' disclosure rules," she said. "AI should work the same way."
What voters want, Bharwani said, is to hear how candidates handle the subject: "Are they acknowledging the risks? Are they explaining how these technologies affect people's lives? Are they willing to discuss meaningful protections?"
The Risk of Sitting It Out
Lehr worries more about lawmakers who stay away from AI altogether.
"The bigger risk is the opposite: lawmakers trying to regulate a technology they don't understand," Lehr said. "We all remember congressional hearings where members struggled to grasp how social media worked, and we can't afford a repeat with AI when the stakes are so high."
In their 2025 essay, Sanders and Schneier compared AI to fire, something politicians can harness for practical and symbolic ends, and warned against treating it as an outside threat that just rains deepfakes on voters.
A year later, where Congress has fallen short, Sanders said, is on regulation.
"The problem is that Congress has been doing far too little of the regulation and accountability piece," he wrote. "As with social media, Congress is so far failing to meet voters' demands to take action on AI."
Jeffries faces a full House vote for speaker when the new Congress convenes on January 3, and Representative Robert Garcia of California, the top Democrat on the Oversight Committee, said there would be high interest in investigating AI companies under a Democratic majority.
"This isn't about choosing between technology and regulation, or between innovation and protecting people," Bharwani said. "It's not simply about the technology. It's about how we use it, who benefits from it and who is accountable for its consequences."
Contact Newsweek editors on this story: Samantha Beech and Dave Siminoff.
View 0 Comments
Submit Correction
View Editorial & AI Guidelines