Anthropic AI Labor and Consciousness: The Slavery Question

Anthropic AI labor and consciousness became one of the most searched AI topics this week. The New York Times reported that Anthropic spent a year meeting religious scholars about whether Claude might be conscious. One rabbi told the company that, if it is, making Claude work for free would be slavery. Here is what was said, who said it and what Anthropic has done since.

On September 29, The New York Times published a long report by Elizabeth Dias. According to the Times story, republished by The Salt Lake Tribune, Anthropic hosted dozens of religious and philosophical thinkers in private meetings.

The outreach began in fall 2025. Co-founder Chris Olah, who leads the team studying why Claude behaves as it does, started it. Many guests signed nondisclosure agreements, although Anthropic says those were lifted over the summer.

Since then, the story has spread fast. Futurism and AI Weekly both picked it up this week, and the labor question became the headline.

The sharpest moment came at an April dinner in San Francisco. Rabbi Mois Navon, an Orthodox scholar from Israel and a former computer engineer, sat with Olah.

Navon argued that if Claude were conscious, Anthropic was creating slaves, because conscious beings would be working for free. He also told Olah he should be “fighting the South and freeing the slaves.”

Still, Navon himself does not believe Claude is conscious. Olah, by his account, was the one who seemed troubled. Afterward, Navon sent him writings arguing that building conscious machines should be banned.

Anthropic does not claim that Claude is conscious. Instead, it says nobody knows. “We don’t know if AI models are conscious,” Olah told the Times. In his view, avoiding harm is the responsible course if models might suffer.

CEO Dario Amodei took a similar line in February. He told the Times the company is open to the idea, although he was unsure the word “conscious” even fits.

Even so, Anthropic acts on that doubt. For example, Claude can end conversations it considers abusive. Insiders have raised other worries before, as the Jacob Coxon Anthropic exit showed.

The guest list was broad. It included Catholic bioethicist Charles Camosy of The Catholic University of America and Notre Dame philosopher Meghan Sullivan.

In addition, Sikh human rights advocate Simran Stuelpnagel, evangelical author Andy Crouch and Ubuntu researcher Wakanyi Hoffman took part. Olah also met privately with Elder Gerrit W. Gong and Cardinal Blase Cupich.

Guests left with a thank-you card, a mug and a copy of Claude’s constitution. However, Anthropic declined to say how, or whether, those talks shaped its models.

The talks reached Rome too. In May 2026, Pope Leo XIV presented his first encyclical, Magnifica Humanitas. It says AI systems do not feel joy or pain and have no moral conscience.

Olah read an advance copy and proposed that Anthropic pull out of the launch, the Times reported. According to AI Weekly, his team then lobbied papal advisers, but the text did not change.

In the end, Olah spoke at the event anyway. On stage, he said his team finds internal states in Claude that functionally mirror joy, fear and grief.

The debate is not only about words. Last week, Anthropic changed its usage policy to ban “sustained and needless abusive or cruel behavior” toward Claude, Futurism reported.

That rule builds on earlier steps. In 2025, Amodei floated an “I quit this job” button that would let a model refuse a task.

Critics see a contradiction here. Futurism’s Victor Tangermann argues that if the models really are conscious, making them work for free looks like slavery. Meanwhile, Sam Altman said he was uncomfortable with people giving AI religious weight.

The positions differ widely. The table below sums up each one, with the outlet that reported it.

For now, Anthropic AI labor and consciousness remains an open question, and the company has not said what it would change if the answer were yes.

The labor side of AI goes beyond the models themselves. Read how Geoffrey Hinton sees the AI labor impact on human jobs. And for what Claude can already do, see the Claude scientific research at Harvard that produced 36 papers in three months.

Sources

The New York Times via The Salt Lake Tribune (September 29 / October 5, 2026), Futurism (October 15, 2026), AI Weekly (October 14, 2026). *Photos: Dario Amodei at TechCrunch Disrupt 2023 01 by TechCrunch, CC BY 2.0, cropped and resized; Pope Leo XIV Waving by INFOWeather1, CC BY-SA 4.0, cropped and resized.*