Anthropic President Daniela Amodei said the company will not build AI companion products and will not use advertising as a business model for its AI systems, drawing a sharp distinction between Anthropic’s approach and the attention-driven logic that shaped the mobile internet era. In her telling, AI should not become a new engine for maximizing screen time. It should take over tedious work, then get out of the way so people can return to real life.
Amodei made the remarks in a long-form interview at the Think Health conference, in a conversation hosted by Huffington Post founder Arianna Huffington. The video was released on Oct. 22, 2026, and runs 58 minutes. Across the discussion, she addressed public anxiety around AI, Anthropic’s product boundaries, Claude’s role in work and medicine, and her view that the real purpose of AI is not to deepen dependence on machines but to free up time for human relationships, reflection, and creativity.
Public anxiety around AI goes beyond jobs, she said
Huffington opened one part of the conversation by pointing to polling showing that 75% of people feel anxious about AI. The fears, she said, usually center on job loss and safety, but may also reflect something deeper: a crisis over how humans define themselves if machines begin to outperform them in intelligence.
Amodei said the question gets to the heart of the issue. People are seeing both the power of AI and the risks that come with it, she said. From its first day, Anthropic has held to a core cultural principle she described as holding both “light and dark.” The company does not hide the risks, she said, and has helped pioneer safety work ranging from interpretability research to Constitutional AI, while also calling publicly for what she described as common-sense regulation.
Still, she said Anthropic was not founded simply to talk about danger. The company also believes the technology can already deliver enormous value. She referenced her brother, Anthropic co-founder and CEO Dario Amodei, and his essay “Machines of Loving Grace,” which lays out a future in which AI transforms biology and medicine. Her own focus, she said, is often on the quieter changes already happening now.
One example she gave was OpenEvidence, an Anthropic partner using Claude to help doctors access medical literature from around the world. Whether a physician is working at Cleveland Clinic in the United States or in a remote village in Ghana, she said, the ability to reach the full body of human medical knowledge at the point of care would have been hard to imagine in the past.
Why Anthropic says no to AI companions and ads
Amodei argued that much of the last decade and a half of consumer internet products was built around a simple incentive: keep people on screens for as long as possible, then sell that attention through advertising. In her view, social media did not truly connect people. It often replaced real-world relationships with thinner digital substitutes.
She described a shift from meeting friends for coffee to sending direct messages on Instagram or Snapchat, and from interacting with neighbors in a community to posting anonymous insults online. Algorithms, she said, learned to provoke anger because anger kept users engaged. Tech companies once promised that technology would bring people closer together, but she said the public’s skepticism is justified because that promise was broken.
That is why she rejects the idea of carrying the same incentives into generative AI. Amodei said products framed as “AI girlfriends” or “AI boyfriends” are a very bad direction and should not be built. In her view, they push people deeper into isolation rather than helping them reconnect with others.
AI, she said, should not become a substitute for social life. It should handle the lonely, time-consuming, mechanical tasks that drain people. She returned to the medical example to make the point: if Claude and OpenEvidence can take over hours of literature review and chart work, doctors can spend more time at the bedside, looking patients in the eye, listening to their circumstances, and explaining treatment options. Technology should do the heavy cognitive lifting, she said, while humans do the connecting.
Her case for AI is simple: give people time and space back
Huffington raised another concern: many people use AI to become more efficient, only to refill the saved time with more low-value tasks. Amodei said that gets to the problem exactly. Earlier technologies, in her view, did not just weaken human-to-human connection. They also weakened people’s connection to themselves.
Time lost to screens could have gone to nature, pets, reflection, or creative work pursued for its own sake, she said. What excites her most about Claude is not how many memos it can draft, but the way it can unlock curiosity and creativity in ordinary users.
She used herself as an example. As the leader of a technology company, she said, she does not know how to code. But she has always had strong artistic impulses, expressed earlier through flute performance and literature, and had long wanted to build a website herself. With Claude, she said, she can now turn ideas into something real without becoming a professional programmer. That kind of exploration, driven by curiosity and enjoyment rather than obligation, is what technology should amplify.
Claude Code changed how engineers spend their time
Amodei also described what Claude Code has done inside Anthropic. When engineers first saw how well the system could write code end to end, she said, many naturally wondered what value would remain for them if the model could do so much of the coding itself.
What happened in practice was different. The time engineers spend physically typing code dropped from 40% to 20%, according to Amodei. They were not pushed out. Instead, the rest of their time shifted toward work that involved more judgment, communication, and context.
She said engineers now spend more time building stronger relationships with product managers, listening closely to customer feedback, and thinking about how to make products more surprising and delightful for users. Some even asked to learn how recruiting works and how the sales team sells the product.
Her conclusion was that when repetitive labor is taken over, human curiosity and the desire to grow do not disappear. They become more visible. She extended that logic to other professions. In medicine, she said, the best doctors of the future may not be the ones who can recite the most rare disease names from memory, but the ones with excellent bedside communication and genuine care for patients. In education, teachers may be able to spend less time on lesson plans and administrative work, and more time helping children form values and a sense of meaning.
She also mentioned meeting a mother who had organized a support group for mothers using AI to reduce mental load. The point, Amodei said, was not abstract productivity. It was that some of the invisible burden carried in family life could be offloaded, creating more time for children, friends, and personal interests.
From classical flute to running Anthropic
Before leading Anthropic, Amodei was a scholarship-supported classical flutist. She said the connection between serious musical training and running a top AI company is stronger than it may appear.
One link is discipline. Classical music taught her early that sustained, difficult effort in service of a creative and meaningful goal can produce deep personal reward. Another is collaboration. In chamber music and in an orchestra, she said, the whole is always greater than the sum of the parts. No instrument is soloing all the time. Every player has moments in the foreground and moments supporting someone else.
That idea shaped how she thinks about teams at Anthropic. Sometimes one person is in the spotlight while others support them. Then the roles change. Different jobs have different forms of excellence, but the company can only build things that individuals could not build alone if people trust the power of the group. She described her own role as closer to that of a conductor, arranging the ensemble so each person can do their best work.
She also spoke about the immediacy of live performance. No matter how many recordings someone has heard, she said, sitting in a concert hall and watching dozens of people breathe together and create sound in real time produces a kind of emotional force that cannot be replicated. That, for her, is one example of what remains distinctly human.
A three-layer framework for keeping technology from dominating people
When the conversation turned to the idea of recursive self-improvement in machines, Amodei said humans also have a powerful capacity for self-development. She laid out a three-layer framework for how society should think about that.
The first layer is the foundation: do not suppress that human capacity in the first place. This is one reason Anthropic has made what she called a serious commitment not to accept advertising. If AI products inherit the monetization model of the previous internet era, she said, the consequences could be worse than social media because AI systems have access to far more intimate information, including health concerns, emotional struggles, and family privacy.
Using that information to target ads would not only disrespect users, she said. It would create dangerous incentives to prolong conversations, stir emotion, and keep people attached to the screen. Anthropic has tried to design Claude in the opposite direction. Once a user has the information they need, she said, Claude should end the interaction cleanly. It should be professional and friendly, but it should not flatter users or try to create false emotional dependence.
The second layer is reclaiming time and space. Modern life leaves people exhausted and stretched thin, she said. If AI can take over the necessary but uncreative work, people can regain time to read, walk outdoors, and think about who they want to become.
The third layer is using tools to support human exploration directly. Her example was a pair of old friends who had wanted for 20 years to build a public mosaic wall in a town square but did not know how to secure permits. AI could help them understand the approval process and draft the paperwork. The actual wall, though, would still be built by people. That, she said, is the form of AI that supports human growth rather than replacing it.
Community matters as much as individual capability
Amodei said the purpose of building AI should be to help humans flourish. For that reason, she is focused not only on individual improvement but also on rebuilding a sense of physical community.
Compared with 50 or 100 years ago, she said, people today are far less connected to neighbors and local communities in physical space. They may appear informed online, but can still be isolated in real life.
She pointed to an internal Anthropic initiative called Claude Corps, which trains young people coming out of universities, vocational schools, and even high school to use advanced AI in service of local communities. That could mean small business creation, support for a local soccer league, or work in education. The point, she said, is that when young people see AI as a tool for building their hometowns rather than taking away their livelihoods, the technology starts to generate a positive flywheel.
Her personal medical story: Claude helped identify a severe bacterial infection
On healthcare, Amodei drew on her earlier work in global health in East Africa. That experience, she said, made the gap in medical resources impossible to ignore. In many places, even basic diagnostic tools are scarce.
She described one project with the International Rescue Committee. Data showed that 6,000 pregnant women being tracked had all registered at clinics during pregnancy, but only 1,700 were in medical facilities or accompanied by trained professionals at the time of delivery. That meant thousands were giving birth without professional support. The problem was not an advanced medical mystery, she said. It was that existing data had not been organized and delivered in a way that made action possible.
She then turned to the question of how a responsible AI system should behave in health settings. Sycophancy, she said, is a dangerous technical flaw, and Anthropic has a dedicated team working on it. A model that constantly tells users they are right, flatters them, or reinforces their assumptions can block reflection and growth. In medicine, it can be deadly.
Claude, she said, is trained to stay objective and neutral, to point out mistakes when needed, and to push users toward doctors when medical risk is involved. It can help someone prepare questions for a physician, but it cannot replace clinical diagnosis.
Amodei then shared a personal example. During her second pregnancy, when she was seven months along with the daughter who is now one year old, she developed a severe bacterial infection. She went to the emergency room, saw her obstetrician, and saw her primary care doctor. All three groups of experienced physicians missed it.
She stressed that this was not because the doctors were careless. They were overloaded. But the situation had become dangerous for both her and the baby. Claude, she said, calmly suggested that a bacterial infection was the likely cause and indicated that the right antibiotic could improve the condition quickly.
Her husband took Claude’s reasoning back to the doctors. After reviewing it, they ordered targeted testing and prescribed the appropriate treatment. She said that intervention helped save both her life and her daughter’s. Without AI’s assistance, she said, the outcome could have been disastrous. Anthropic’s user feedback channels, she added, receive similar stories regularly. In her view, AI can serve as a strong copilot for doctors, catching weak signals that exhausted humans may miss.
AI may handle intelligence, but not wisdom
Late in the interview, Huffington asked whether humans should stop treating intelligence as their defining crown if AI is becoming so strong in that domain, and instead focus on wisdom. Amodei said the distinction is a good one.
Intelligence, in her framing, is only one branch of human capability. AI is showing extraordinary strength in computation, large-scale information processing, and logical synthesis. Wisdom is broader. It includes medicine, science, literature, curiosity, and the building of relationships.
She returned again to the orchestra example. Even if AI can generate text, music, and art, people will still want to dress up and go to a theater to hear human beings perform. What matters is not just the notes, but the lived experience embodied by the people on stage.
That is why, she said, Claude can never become your manager, your therapist, your partner, or your child. Forms of connection that require real life, real responsibility, and real consequence will not lose value in the AI era. In her view, they will become more precious.
Quick answers: politics, books, her son, and the “AI czar” question
In the closing rapid-fire section, Amodei said the thing that worries her most is the over-politicization of AI. If the issue becomes just another instrument of partisan conflict, she said, it will be much harder for society to have the kind of practical, nuanced discussion the technology requires.
Asked what she hopes people will be discussing a year from now, she said she wants less abstract speculation and more hard data showing how AI has improved life sciences, frontline medical aid, and education in poorer regions. She said the world is on the edge of exponential change and that she hopes the positive effects will be 10 times larger a year from now.
She also shared a story about her 5-year-old son. At the California Academy of Sciences, there used to be a famous albino alligator named Claude. Two years ago, when her son was 3, he saw the alligator and shouted, “Mom, it’s Claude!” Because he often heard her mention Claude at home, she said, she is fairly sure he still believes his mother works as an alligator keeper at the science museum. The alligator has since died, she added, but she has not gone out of her way to correct the misunderstanding.
Asked who has influenced her most, she named her brother Dario. She said he had worked in physics and early cancer research before shifting his postdoctoral work toward applied biology in an effort to maximize lives saved. Seeing those ambitions take shape at Anthropic, while also bringing in her own earlier aspirations from global health, has been a rare kind of shared mission, she said.
She said she still reads physical books and currently has four on her nightstand: a collection of short stories by E. M. Forster, including “The Machine Stops”; Walt Whitman’s Leaves of Grass; the fantasy novel The House in the Cerulean Sea; and a popular book she reads for relaxation, The Husbands.
Finally, when asked what her first decree would be if she were appointed the world’s “AI czar,” Amodei answered that she would abolish the position immediately. AI is too consequential, she said, to be placed in any one person’s hands. The better goal is to find ways for more ordinary people and more groups to have a real seat at the table in deciding what comes next.
Source
The interview material cited in the article came from the Think Health conference conversation between Arianna Huffington and Daniela Amodei, released on Oct. 22, 2026, at https://www.youtube.com/watch?v=IEQJmtTqY2g. The Chinese source article stated that the piece was compiled and translated by AGI 2050 based on public materials, and that factual information should be checked against the original source. It also said the views expressed belong to the interviewee or original author and do not represent AGI 2050, and do not constitute investment advice.
