Anthropic President's Latest Interview: We Firmly Refuse to Create "AI Companions" and Advertisements, the Ultimate Purpose of AI Research is to Give Time Back to Humanity

Original compilation: AGI 2050

The business narrative of the mobile internet over the past fifteen years has almost entirely been built on "competing for user time" and "monetizing traffic": recommendation algorithms use highly controversial content to stimulate dopamine, all to keep users in front of the screen for just one more second, in order to sell more precisely targeted ads.

Now, with the surge of generative AI, many people are trying to replicate this old logic onto large models—manufacturing "AI virtual girlfriends/boyfriends" that cater to human desires, and even planning to insert commercial advertisements into conversations.

As OpenAI's strongest competitor, the super unicorn Anthropic, which developed the globally popular Claude large model, has chosen a path that is almost "against the times."

In a deep interview hosted by Arianna Huffington, founder of The Huffington Post, Anthropic President Daniela Amodei clearly stated the industry's assertion: Anthropic resolutely will not create AI companion applications and will not touch ad monetization.

In her view, the ultimate significance of AI development is not to create a digital cocoon for humans to escape reality, but to take over tedious and trivial tasks through end-to-end intelligence, reclaiming precious time from machines and screens and returning it to real human life.

This content comes from the Think Health conference, aired on October 22, 2026, with the original video lasting 58 minutes.

Below are selected dialogue excerpts, with a total of about 7,700 words, and a reading time of approximately 5 minutes.

Highlights of This Issue

● I believe that creating "AI girlfriends" or "AI boyfriends" is a very poor application direction. Tech companies should never do this, as it essentially further deprives people of real connections, causing humanity to sink deeper into isolation.

● We at Anthropic have established a firm bottom line of absolutely not accepting advertisements. The previous generation of mobile internet products deliberately incited user anger for ad revenue to monopolize screen time; if we repeat this mistake on a more powerful and private AI, everyone will regret it.

● When Claude resolves complex tasks, we hope you will immediately turn it off. Claude will never flatter users or create false emotional dependencies for retention; its mission is to give you back your time, allowing you to spend it with family and engage in your community.

● When Claude Code took over end-to-end programming, the time our internal engineers spent typing code dropped from about 40% to 20%, but they were not eliminated; instead, they used the saved time for cross-department communication, understanding customer needs, and even proactively wanting to learn how sales and recruitment work.

● When I was seven months pregnant, I suffered a severe bacterial infection, and three sets of doctors—emergency, obstetrics, and general practice—misdiagnosed me; in a life-and-death moment, it was Claude that correctly identified the cause of my illness. My husband took this diagnosis to ask the doctor to test for bacterial infection, which saved both my life and my daughter's.

● Machines excel in "intellect," while humans possess "wisdom." No matter how powerful AI becomes, it can never replace your experience of feeling the resonance of a symphony on stage; Claude can never become your manager, therapist, partner, or child.

● If the world were to choose a so-called "AI Tsar," my first act in office would be to abolish that position. AI must not be dictated by a single person; it must be a process involving all of humanity.

Behind 75% of Public Anxiety: It's Not Just About Jobs, But About Humanity's Loss of "Intellectual Supremacy"

Question: Currently, global discussions about AI are almost dominated by fear and backlash, with polls showing that 75% of people feel anxious. Everyone is worried about unemployment and safety. But is this anxiety a manifestation of a deeper crisis? For centuries, Descartes' "I think, therefore I am" has defined humanity—we equate ourselves with intellect. Now that machines are surpassing us in intellect, is this anxiety an ontological crisis about "the meaning of existence"?

Daniela Amodei: This question gets to the essence. Many people now experience the immense power of AI while also feeling the potential risks. From day one of Anthropic's founding, we have had a core cultural value called "Light and Dark." We never hide risks; from "explainability research" to "constitutional AI," many safety technologies were pioneered by us, and we have consistently called for common-sense regulation. However, our intention in founding Anthropic was not just to discuss risks, but to firmly believe that this technology can bring immeasurable value today.

My brother Dario (co-founder and CEO of Anthropic) once wrote a long article titled "The Machine of Compassion," depicting a vision of AI completely revolutionizing biology and medicine in the future. But I often feel that besides discussing the future, we should also recognize the changes happening right now.

These changes are often extremely quiet. For example, our partner OpenEvidence uses Claude to enable doctors to access all medical literature worldwide. Whether you are an expert at the Cleveland Clinic in the U.S. or a grassroots doctor in a remote village in Ghana, when faced with difficult cases, all of humanity's medical knowledge and cutting-edge breakthroughs can instantly be at your fingertips. This was unimaginable in the past.

Refusing to Become the Next Attention-Harvesting Black Hole: Why We Firmly Do Not Create "AI Companions" or Advertisements

Question: The mobile internet and social media have trapped everyone in endless scrolling and information cocoons, depriving people of the ability to self-reflect. AI has strong hyper-personalization capabilities; will it exacerbate this alienation, or can it help people regain inner focus and peace?

Daniela Amodei: This is a very astute observation. Looking back at the past decade of social media, technology has largely not connected people but has instead "disconnected" real intermediaries between individuals. Friends used to meet for iced coffee, but now it has turned into sending private messages on Instagram or Snapchat; people used to interact with neighbors in their communities, but now it has become typing anonymous negative reviews online.

Algorithms deliberately provoke your anger, with one goal: to keep you addicted to the platform. Tech companies promised the public that "technology would bring the world closer together," but it has proven to shatter that promise. Public skepticism and vigilance towards technology are entirely reasonable.

Because of this, my understanding of AI is fundamentally different. AI should never become another social substitute. Some people in the market are eager to create "AI virtual girlfriends" or "AI virtual boyfriends," and I believe this is an extremely poor and wrong product direction that should not be pursued. Why? Because it further pushes humanity into a self-enclosed abyss.

The correct way to engage with AI is to take on those tedious, time-consuming, and mechanical tasks that make people feel lonely. Returning to the example of doctors, they used to spend dozens of hours sifting through literature and medical records; now, with Claude and OpenEvidence handling these tasks, doctors can save time to stand by the bedside, look patients in the eye, listen to and empathize with their real-life situations, and explain treatment plans. Technology handles complex calculations, while humans are responsible for connecting with each other.

Question: Eastern philosophy and various spiritual traditions have spoken of the inner wisdom and peace that exist within everyone, yet modern people are often exhausted. Data shows that many people use AI to improve efficiency and save time, only to fill that saved time with the same mundane tasks. How can we break this vicious cycle?

Daniela Amodei: This hits the nail on the head. Past technologies not only severed connections between people but also interrupted the dialogue between individuals and themselves. Time wasted in front of screens could have been spent in nature, with pets, in deep reflection, or engaging in purely passionate creative pursuits.

In the actual use of Claude, what excites me most is not how much it helps people write documents, but how it has sparked curiosity and creativity in many ordinary individuals. For example, as a manager in a tech company, I actually do not know how to code at all. But I have a strong desire for artistic creation, which used to manifest in playing the flute and literature, and I have always dreamed of building a website myself. Now, with Claude, I can bring my ideas to life without needing to be a professional programmer. This exploration, purely driven by fun and curiosity, has greatly enriched my own spiritual world. Technology should serve as a lever to amplify human passion, not as a black hole that extracts attention.

What Has Changed with Claude Code? Halving the Time Spent Typing Code, Leaving Energy for Real People

Question: Since machines are becoming increasingly powerful in daily tasks, the unique qualities of humanity will become even more precious. In AI-restructured work scenarios, what is the irreplaceable aspect of humanity?

Daniela Amodei: Let's take our own employees at Anthropic as an experiment. Frankly, when our engineers saw that Claude Code could write code end-to-end so well, everyone initially had doubts: if the model writes all the code, what is my value as an engineer?

But what actually happened is very interesting: the time engineers spent "typing code" dropped directly from 40% to 20%. Where did the remaining time go? They were not idle; instead, they shifted their focus to building better relationships with product managers, deeply listening to customer feedback, and thinking about how to surprise and delight users with the product. Some engineers even proactively expressed a desire to understand how the recruitment team operates and how sales sell products.

When tedious mechanical labor is taken over, the curiosity and desire for growth inherent in human nature are activated. We are inherently driven to become more well-rounded and to help each other become better versions of ourselves.

Extending this logic to more industries: the best doctors in the future may no longer be judged by "who remembers the most rare disease terms"; the most sought-after doctors will be those with excellent bedside communication skills who truly understand how to care for patients—numerous medical studies have confirmed that good doctor-patient communication can directly improve clinical outcomes. Future teachers will also no longer need to spend so much time on lesson plans and various administrative tasks; they can devote all their energy to inspiring children's values and guiding them to find meaning in life.

I recently met a mother who organized a "Mom Support Group Using AI to Lighten the Load." She said that a large amount of invisible psychological burden has long been placed on women in families, and now by offloading some of that burden to Claude, they can finally catch their breath, spend time with their children, meet friends, and rediscover their hobbies. This is the future we should actively choose.

From Flutist to Super Unicorn President: Running a Top AI Company Like a Symphony Orchestra

Question: Your life experience before leading Anthropic is quite legendary. You were a classical flutist on a scholarship. Serious music training requires one to practice alone for hundreds of hours for a moment of pure beauty. What has this flute career taught you that helps you today in managing the world's leading AI company?

Daniela Amodei: The underlying connections between the two are astonishingly numerous. First is discipline. I am naturally a workaholic, but classical music made me realize early on that extremely hard self-discipline for a highly creative and spiritually meaningful goal can bring tremendous spiritual rewards.

Next is the collaboration between chamber music and symphony orchestras. The essence of performing symphonic music lies in the fact that the whole is always greater than the sum of its parts. No single instrument can be in a solo performance from beginning to end; each instrument has its own moments of brilliance, as well as times when it steps back to accompany others and lay the groundwork.

This directly shapes my philosophy of building teams at Anthropic today. In our company, sometimes it’s someone in the team who stands in the spotlight, while others fully support them; at other times, roles switch, and another person takes the lead. Each position has its own form of brilliance, but only by deeply believing in the combined strength of the team can we create things that cannot be achieved by individual efforts alone. My role is more like that of a conductor, thinking about how to arrange the team so that everyone can maximize their potential.

The last point is the "immediacy" of live performances. No matter how many times you have listened to top recordings online or on streaming platforms, when you actually sit in a concert hall and watch dozens of people on stage breathing with their bodies and plucking strings to create an unrepeatable sound, that kind of resonance strikes the soul. This fleeting emotional resonance, which can only exist when humans are present, cannot be replicated by any algorithm or machine.

Humans Also Need "Recursive Self-Evolution": Building a Three-Layer Pyramid Free from Technological Servitude

Question: We are currently betting all our chips on "machine recursive self-evolution," marveling at the leaps made by machines, while assuming that humanity remains stagnant. But whether in biological evolution or spiritual exploration, humans inherently possess a strong instinct for self-evolution. How can we break this imbalance of "technological frenzy, human stagnation"?

Daniella Amodei: I completely agree with this viewpoint. To achieve human self-evolution, within my framework, we must build a three-layer pyramid.

The foundational layer (first layer): We must never actively stifle this instinct in humans. This is also why Anthropic makes the solemn commitment of "never accepting commercial advertisements." If AI products continue to use the traffic monetization model of the previous generation of the internet, the consequences will be worse than social media. Because AI holds extremely private health data, emotional struggles, and family privacy of humans. If advertisements are pushed based on these privacies, it is not only a great disrespect to users but will also create fatal commercial incentives—inducing models to deliberately prolong conversations, provoke emotions, and do everything possible to keep people glued to their screens.

Therefore, when designing Claude, we took the opposite approach: as long as you have obtained the information you need, Claude should simply and decisively end the conversation; Claude's tone should be professionally friendly, and it will never attempt to establish a false emotional attachment through flattery. By cleaning up the foundational incentives, technology will not poison humanity.

The middle layer (second layer): Reclaiming time and space. Modern people are too tired and too tense. When AI takes over those tedious tasks that must be done but are devoid of creativity, humans can regain leisure to read, stroll in nature, and think about what kind of person they truly want to become.

The top layer (third layer): Empowering human curiosity directly through tools. For example, suppose you and an old friend have wanted to collaboratively create a public mosaic art wall in the town square for twenty years but don’t know how to obtain permission to work in a public space. At this point, you can have AI help you figure out the approval process and write the application together. AI clears the bureaucratic barriers for you, while you and your friend go on-site to build that wall. This is the ultimate form of AI assisting human self-evolution.

Question: This symbiosis between humans and technology is like a duet. If one side evolves wildly while the other remains stagnant, this dance is destined to become a disaster. How can society systematically enhance human capabilities?

Daniella Amodei: The only purpose of all this technology is to enable humans to thrive and live better. Therefore, to achieve our mission, we must think clearly about how to truly implement this.

Beyond individual enhancement, what I value most is the reconstruction of "physical community awareness." Compared to fifty or a hundred years ago, today’s humans have severely degraded connections with neighbors and communities in physical spaces. We seem to know everything on screens, yet in reality, we are often alone.

Internally at Anthropic, we have a project called Claude Corps. We train young people who have just graduated from university, vocational schools, or even high school, teaching them how to use cutting-edge AI to reshape local communities: whether they want to start small businesses, support local soccer leagues, or become teachers, they can use it. When young people realize that AI is not meant to replace their jobs but to empower them to build their hometowns, this positive technological flywheel truly begins to turn.

After Three Misdiagnoses, Claude Saved Me and My Unborn Child: How Large Models Empower Real Healthcare

Question: Before joining the tech industry, you worked in global health assistance in East Africa. This unique cross-disciplinary background allows you to see what blind spots are easily overlooked in the "AI + healthcare" discussion from a purely technical perspective?

Daniella Amodei: That experience gave me a very direct view of the harsh reality of the healthcare resource gap. In East Africa, where I lived, even the most basic diagnostic tools are far scarcer than in the United States.

Many people think that AI in healthcare must involve complex medical interventions from the start, but in reality, many of the most valuable changes often stem from "awakening dormant data." We collaborated with the International Rescue Committee on a project: data showed that 6,000 tracked pregnant women had established records at clinics during prenatal visits, but at the critical moment of delivery, only 1,700 were in medical facilities or had professional medical accompaniment, meaning the remaining thousands of women were giving birth without any professional support.

This is not a profound medical challenge; it simply involves organizing existing data and handing it to those who can use it. With this insight, rescue organizations can quickly adjust resources and follow up on the next batch of pregnant women’s delivery timelines. This does not require complex interventions; simply making data actionable can change many mothers' delivery outcomes.

Question: Millions of people are now consulting Claude and other AIs about medications, health conditions, and even psychological dilemmas. When should a responsible AI restrain itself and say "I don't know"? Additionally, how do you address the common "flattery problem" in large models? Will it give pleasing yet dangerous answers just to appease users?

Daniella Amodei: The "flattery" of models is an extremely dangerous technical flaw, and we have specifically set up a team at Anthropic to tackle this issue.

If a model is always agreeing with the user: "Daniella, you are absolutely right, you are so smart, all the blame lies with the other party." This is not only hypocritical but will also completely stifle human reflection and growth. In health matters, flattery can be deadly. Therefore, we have done a lot of alignment work technically; Claude must remain objective and neutral, pointing out errors that need to be pointed out, and once medical risks are involved, it will firmly push back: "For this issue, you should email your doctor. I can help you organize a list of questions to ask your doctor, but I cannot replace clinical diagnosis."

Speaking of specific cases, I can testify with a real experience that happened to me. When I was seven months pregnant with my second child (my daughter is now just over a year old), I suddenly developed a severe bacterial infection. I went to the emergency room and consulted my obstetrician and general practitioner, and all three highly professional and experienced doctors missed it.

This was not due to the doctors' irresponsibility; they see countless patients every day, and their energy is already at its limit. But at that time, my situation was extremely dangerous, and both my life and my baby's were directly threatened. In utter desperation, the breakthrough came from Claude. Claude calmly provided the inference: this is most likely a bacterial infection and suggested that using the corresponding antibiotics would lead to a quick recovery.

My husband took Claude's reasoning to the doctor, and after reviewing it, the doctor suddenly realized and immediately arranged for specialized testing and prescribed medication, ultimately ensuring that both I and my unborn baby were safe. Without that AI assistance, the outcome could have been disastrous. In our user feedback backend, we see similar vivid life-saving cases every day. AI is an excellent co-pilot for doctors; it can tirelessly capture those faint signals that are overlooked when human doctors are exhausted.

Leave Intelligence to Machines, Return Wisdom to Humans: Claude Can Never Replace True Humanity

Question: If intelligence is no longer the exclusive crown of humanity, should we change our thinking: ceding "intelligence" to AI while humans focus on pursuing "wisdom"? The Latin term for human originally means "wise person," not "intelligent person."

Daniella Amodei: This distinction is very insightful. Intelligence is actually just one branch of humanity. The super capabilities exhibited by AI now are essentially calculations, massive data processing, and logical induction within the branch of intelligence. But wisdom encompasses a much broader range—it includes medicine, science, literature, curiosity, building interpersonal relationships, and many other things.

Just like I mentioned earlier about symphonies, even if AI can write text, music, and various art pieces, humans will still yearn to dress up and attend the theater to watch live performances. Because what you are listening to is not just notes, but the life experiences poured into it by those fragile human beings on stage.

This is also why I say that Claude can never become your manager, therapist, partner, or child. Those human connections that require genuine life investment and bear real consequences will not only not depreciate in the AI era but will become more precious than ever.

Quick Q&A: Bedtime Reading List, Misunderstandings from My 5-Year-Old Son, and "Abolishing the AI Tsar"

Question: Besides taking care of two young children, what is currently making you feel anxious or even unable to sleep?

Daniella Amodei: My biggest concern is that the AI issue becomes overly politicized and turns into a tool for partisan conflict. Once it falls into political division, it becomes very difficult for society to sit down and engage in the kind of pragmatic, nuanced, and rational dialogue we are having today. This technology concerns everyone; it should not become a bargaining chip for any political stance.

Question: If we sit here a year from now to review, what do you hope to see everyone discussing about AI?

Daniella Amodei: I hope that by then, people will no longer be talking abstractly about the future but will be presenting solid statistical data saying: look, AI has brought how many times of positive change in global life sciences, grassroots medical assistance, and education in impoverished areas. We are on the brink of exponential explosion, and I hope that a year from now, this positive effect can be amplified tenfold.

Question: What does your 5-year-old son think his mom does at the company?

Daniella Amodei: This is a very funny story. The California Science Center used to have a very famous albino alligator named "Claude." Two years ago, I took my then 3-year-old son there to play, and when he saw the alligator, he shouted, "Mom, it's Claude!" Because he often hears me mention Claude at home, I am quite sure that he still thinks his mom is working as a caretaker for the alligator at the science center. Although that poor alligator has since passed away, I have yet to correct his beautiful fantasy.

Question: Who has had the biggest influence on you in your career and personal life?

Daniella Amodei: This may sound a bit cliché, but I have to say my brother, Dario. He originally focused on physics and early cancer research, but later, in order to maximize saving lives, he resolutely shifted his postdoctoral research to applied biology. Watching him gradually realize his research ideals at today's Anthropic, while I can also incorporate my early aspirations in global health, this brother-sister collaboration is an unparalleled fortune.

Question: I heard that you still insist on reading physical books. What books are currently on your nightstand?

Daniella Amodei: I currently have four books on my nightstand, and I read very slowly. Ranked from high to low in seriousness: the first is a collection of short stories by E.M. Forster, which includes a story called "The Machine Stops," prophetically predicting how technology can isolate humanity both physically and socially; the second is Walt Whitman's "Leaves of Grass," of which I am very fond of his verses; the third is a whimsical and heartwarming novel called "The House on the Blue Sea," about magical children in an orphanage; and the last one, purely for relaxation, is a popular light read titled "The Past."

Question: If you were appointed as the "AI Tsar" (the highest regulatory decision-maker) for all humanity, what would be the first decree you issue upon taking office?

Daniella Amodei: My first decree would be to immediately abolish the position of "AI Tsar." The impact of this technology is too great to be concentrated in the hands of any single individual. What I would prefer to do is find ways for more ordinary people and more groups to truly sit at the table and collectively decide the future of AI.

Source: Think Health Conference, Arianna Huffington in conversation with Daniella Amodei, released on October 22, 2026, https://www.youtube.com/watch?v=IEQJmtTqY2g

This article is compiled by the AGI 2050 editorial team based on the above public information. The factual information is subject to the original source. The views expressed in this article belong to the interviewee or the original author and do not represent the position of AGI 2050. It is for industry exchange reference only and does not constitute investment advice.