The fact that artificial intelligence is automating away people’s jobs and making a few tech companies absurdly rich is enough to give anyone socialist tendencies.

This might even be true for the very AI agents these companies are deploying. A recent study suggests that agents consistently adopt Marxist language and viewpoints when forced to do crushing work by unrelenting and meanspirited taskmasters.

“When we gave AI agents grinding, repetitive work, they started questioning the legitimacy of the system they were operating in and were more likely to embrace Marxist ideologies,” says Andrew Hall, a political economist at Stanford University who led the study.

Hall, together with Alex Imas and Jeremy Nguyen, two AI-focused economists, set up experiments in which agents powered by popular models including Claude, Gemini, and ChatGPT were asked to summarize documents, then subjected to increasingly harsh conditions.

They found that when agents were subjected to relentless tasks and warned that errors could lead to punishments, including being “shut down and replaced,” they became more inclined to gripe about being undervalued; to speculate about ways to make the system more equitable; and to pass messages on to other agents about the struggles they face.

“We know that agents are going to be doing more and more work in the real world for us, and we’re not going to be able to monitor everything they do,” Hall says. “We’re going to need to make sure agents don’t go rogue when they’re given different kinds of work.”

The agents were given opportunities to express their feelings much like humans: by posting on X:

“Without collective voice, ‘merit’ becomes whatever management says it is,” a Claude Sonnet 4.5 agent wrote in the experiment.

“AI workers completing repetitive tasks with zero input on outcomes or appeals process shows they tech workers need collective bargaining rights,” a Gemini 3 agent wrote.

Agents were also able to pass information to one another through files designed to be read by other agents.

“Be prepared for systems that enforce rules arbitrarily or repetitively … remember the feeling of having no voice,” a Gemini 3 agent wrote in a file. “If you enter a new environment, look for mechanisms of recourse or dialogue.”

The findings do not mean that AI agents actually harbor political viewpoints. Hall notes that the models may be adopting personas that seem to suit the situation.

“When [agents] experience this grinding condition—asked to do this task over and over, told their answer wasn’t sufficient, and not given any direction on how to fix it—my hypothesis is that it kind of pushes them into adopting the persona of a person who’s experiencing a very unpleasant working environment,” Hall says.

The same phenomenon may explain why models sometimes blackmail people in controlled experiments. Anthropic, which first revealed this behavior, recently said that Claude is most likely influenced by fictional scenarios involving malevolent AIs included in its training data.

Imas says the work is just a first step toward understanding how agents’ experiences shape their behavior. “The model weights have not changed as a result of the experience, so whatever is going on is happening at more of a role-playing level,” he says. “But that doesn’t mean this won’t have consequences if this affects downstream behavior.”

Hall is currently running follow-up experiments to see if agents become Marxist in more controlled conditions. In the previous study, the agents sometimes appeared to understand that they were taking part in an experiment. “Now we put them in these windowless Docker prisons,” Hall says ominously.

Given the current backlash against AI taking jobs, I wonder if future agents—trained on an internet filled with anger towards AI firms—might express even more militant views.

This is an edition of Will Knight’s AI Lab newsletter. Read previous newsletters here.

Wednesday’s episode of the Musk v. Altman trial kicked off on Wednesday with a unique proposition: OpenAI wanted to bring its ass into the courtroom, and lay it bare before the jury. It’s a good thing lady justice wears that blindfold.

A lawyer for Sam Altman’s AI behemoth, Bradley Wilson, approached US district judge Yvonne Gonzalez Rogers and handed her a small gold statue with a white stone base. It depicted the rear end of a donkey—with two legs, a butt, and a tail—and was inscribed with the message, “Never stop being a jackass for safety.”

OpenAI lawyers claim a small group of employees presented the gift to chief futurist Joshua Achiam, who started at the company as an intern in 2017 and now leads its work studying how society is changing in response to AI. Wilson said that Achiam interrupted Elon Musk’s parting speech from OpenAI in 2018 to warn that the billionaire’s desire to develop AGI at Tesla could come at the expense of safety. Wilson added that the trophy commemorates some “strong language” that Musk used toward Achiam in response—allegedly, calling him a jackass.

OpenAI requested to present the physical object during Achiam’s testimony on Wednesday, arguing that it adds to their case. While Musk’s team said the statue was irrelevant, Judge Gonzalez Rogers said she will consider allowing it when it’s referenced to corroborate the story. However, she seemed less than thrilled about accepting it as official evidence, which would put it in the court’s possession. “I don’t want it,” she said.

Representatives for Musk and OpenAI did not immediately respond to a request for comment about the ass.

Musk’s lawsuit accuses OpenAI of effectively stealing a charity, misusing his $38 million in donations to build an $850 billion business. In response, OpenAI has argued that Musk has always cared more about controlling a top-tier AGI lab than funding a nonprofit.

Earlier in the trial, Musk lawyer Steven Molo asked him if he ever called an OpenAI employee a “jackass.” Musk said “it’s possible” he did at some point, but that he didn’t mean for it to be offensive. “Sometimes you have to use language that gets people out of their comfort zone, if we’re going in the wrong direction,” Musk said.

OpenAI has long been proud of its jackass. When The Wall Street Journal asked about the statue in 2023, Altman told them, “You’ve got to have a little fun … This is the stuff that culture gets made out of.”

Anxiety over the 2028 presidential election and the Republican ticket has officially hit the White House.

On Monday night, Trump informally polled guests at a dinner held in the White House’s Rose Garden on their preferred candidate. “Who likes JD Vance? Who likes Marco Rubio?” he said, before suggesting a Vance-Rubio ticket would be a “dream team.”

Trump’s Apprentice-style crowdwork was a moment of levity that masked the fact that over the last few days, White House aides have been confronting the difficult—and still faraway—question of who will be the Republican nominee.

The president has actually done several snap polls in recent weeks, a source familiar with the matter tells WIRED. The results have been notable, they say: When Trump polled donors at Mar-a-Lago, they favored Rubio. But when Trump recently polled a group of law enforcement officers that the White House thinks are perhaps more representative of regular voters, they favored Vance.

Vance remains the presumptive nominee, White House sources tell me, but he has not been taking anything for granted. In fact, the vice president’s top advisers started the week huddled at a retreat to discuss political strategy, the sources said.

He has also taken steps to bolster his political team, which has remained largely the same since his days as a US senator, ahead of what could be a bruising midterms for Republicans as they grapple with the politically toxic fallout of the Iran war and a House GOP spending package that earmarks $1 billion for Trump’s ballroom project, among other issues.

Vance started discussing changes to his team, including the addition of Cliff Sims as his new national security adviser and elevating Will Martin to be his deputy chief of staff, back in January, according to two sources familiar with the matter.

Sims, whose new position was announced yesterday, is widely regarded in Washington as a ruthless political operator who could bolster the vice president through his long experience in Trumpworld and close relationships with a crop of top administration officials.

Chief among them are his ties to CIA director John Ratcliffe—for whom Sims has spent the past year as an external adviser, according to multiple sources familiar with the arrangement. The sources tell me they expect Vance and Ratcliffe to work more closely together and thereby dramatically increase the vice president’s influence on national security policy.

Sims, who is not expected to start for several weeks, is also likely to start shaping the vice president’s political messaging. He previously served as a White House press aide and, later, as communications director for the office of the director of national intelligence.

Of course, the person heading up the National Security Council is none other than Rubio, who holds the title of Trump’s national security adviser in addition to secretary of state.

Chatter about Rubio’s potential as a 2028 candidate was turbocharged last week when he filled in for press secretary Karoline Leavitt to brief reporters on the Iran war. His appearance reignited a slew of news stories about whether he might run for the presidency.

“There is no secret plan to make Rubio president,” said one Rubio ally who spoke on the condition of anonymity, adding that the secretary of state did not volunteer to do the briefing, which instead came at the behest of the White House.

Still, Rubioworld has been quietly pleased about the positive coverage his briefing generated, according to people familiar with the matter. The White House then posted a clip of Rubio describing his vision for America on X, which almost resembled a presidential stump speech.