By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.

Stanford Graduate School of Business

Daniela Amodei, Co-Founder and President of Anthropic: Building AI the Right Way

In this View From The Top interview, Gintare Zukauskaite, MBA ’26, speaks with Daniela Amodei, Co-Founder and President of Anthropic.

Gintare ZukauskaitehostDaniela Amodeiguest

May 8, 202648mWatch on YouTube

Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.

CHAPTERS

Jump to publisher timestamps

Chapter summaries

Welcome, audience poll on Claude, and setting the stage
The host welcomes Daniela Amodei to Stanford’s View From The Top and gauges how many audience members use Claude. Daniela reacts playfully and the host frames her unconventional path into building a major AI company.
- •Host welcomes Daniela to Stanford and the series
- •Audience is asked who uses Claude; Daniela responds with humor
- •Host highlights that Daniela and her brother built a major AI company despite not planning for it
- •Host previews Daniela’s arts and politics background as a starting point
From English literature and politics to seeking impact
Daniela describes her early career as interest- and impact-driven rather than a linear plan. She explains how a desire to “make the world better” led her through international development, global health, and political work before returning to Silicon Valley.
- •Daniela frames her decisions as following what was interesting and high-impact
- •She recounts graduating in 2009 and feeling pressure to find “skills”
- •She describes early motivation around fairness and access to essentials like food, water, and medicine
- •She outlines work in international development, Capitol Hill, and a campaign before tech
Becoming a generalist: curiosity, learning across disciplines, and startup reality
Asked how she moves across fields, Daniela identifies as a generalist who leans on curiosity, learning, and a desire to be helpful. She connects her campaign experience to Silicon Valley’s startup environment and the drive to pursue meaningful work.
- •Daniela calls curiosity and cross-disciplinary learning an underrated strength
- •She emphasizes impact as a constant across changing roles
- •She describes campaigns as intense but world-changing work by small teams
- •She notes startups can enable rapid impact with more resources than politics
- •She links sustained effort to caring deeply about the mission
Learning AI culture at OpenAI: asking questions and knowing your lane
Daniela explains how she learned to navigate technical AI conversations after joining OpenAI. She credits prior experience working with engineers at Stripe, exposure to technically oriented family members, and persistent question-asking, while also focusing on her comparative advantage.
- •Stripe experience provided grounding in working with engineers
- •She says growing up around a tech-oriented sibling reduced fear of technology
- •She learned by asking many questions and relying on patient colleagues
- •She highlights understanding comparative advantage: what she could and couldn’t do
- •She describes success as a mix of curiosity and interpersonal skills
Why Anthropic: building a values-led organization and choosing a PBC structure
Daniela recounts leaving OpenAI with six colleagues to create Anthropic. She frames the move as “running towards” a vision centered on safety and responsibility, including the decision to incorporate as a public benefit corporation while still pursuing commercial success.
- •Daniela says seven people left together and aligned on integrity and impact
- •She frames the decision as pursuing a vision rather than escaping a situation
- •Safety and responsibility are described as central, not peripheral
- •She explains the intent to be commercial while doing AI “the right way”
- •She notes the public benefit corporation structure as a values expression
Choosing co-founders: conflict tolerance, shared vision, and pressure-testing relationships
Daniela describes why her co-founder group works well: long-standing relationships, practiced feedback loops, and an aligned picture of what they are building. She stresses the importance of how co-founders handle conflict and suggests pressure-testing compatibility before committing.
- •She emphasizes conflict navigation as more important than many expect
- •She cites her lifelong conflict-and-repair pattern with her brother as an asset
- •She describes long prior histories among co-founders and overlapping work relationships
- •She stresses alignment on the end vision—co-founders shouldn’t be building different “animals”
- •She suggests testing relationships in real-life situations (like travel) before starting a company
What “AI safety” means at Anthropic: radical responsibility and anticipating harms
Daniela defines AI safety as taking radical responsibility for downstream consequences and learning from prior tech industry failures. She lists a wide spectrum of risks Anthropic works on, from catastrophic misuse to user wellbeing and misinformation.
- •She says “AI safety” is overloaded and offers Anthropic’s framing of responsibility
- •She uses social media as an analogy for unintended externalities at scale
- •She emphasizes planning for worlds where things go right and where things go wrong
- •She names serious misuse risks like chemical/biological weapons and cyber warfare
- •She also cites user wellness, child safety, misinformation, and election integrity as safety work
Safety vs revenue: why they usually align, and where timing creates tension
The host asks about balancing safety with generating revenue, and Daniela argues the conflict is less frequent than assumed. She says businesses want reliability and predictability, but rapidly advancing model capability can force delays when risks aren’t fully understood.
- •Daniela says business customers generally don’t want unsafe or unreliable models
- •She claims safety historically aligned with business incentives due to risk aversion
- •She identifies the main tension as time—needing more confidence and understanding of risks
- •She describes delaying broader access to a new model class to address cyber-related concerns
- •She characterizes withholding releases as uncomfortable but mission-driven
Jobs and AI: mostly complementary today, but roles will change shape
Daniela addresses fears about job loss by describing job evolution rather than simple replacement. She references Anthropic’s economic index as suggesting mostly complementary use so far, while acknowledging some roles may disappear and many will transform.
- •She says AI will change which jobs exist and what they entail
- •She cites their economic index as showing primarily complementary use today
- •She mentions only a small number of replacement cases, especially customer service
- •She discusses uncertainty about future job shapes and timelines
- •She predicts software developers will still exist but spend less time writing code and more on broader work
Preparing society: transparency, experimentation, and political/social readiness
Daniela outlines what she believes needs to happen so people feel less anxious about AI. She calls for humility and research, early and transparent public conversations, experimentation in how work and meaning connect, and broader engagement by government and civil society.
- •She argues for humility: admitting uncertainty while doing research
- •She frames publishing the economic index as a way to ground public understanding
- •She calls for starting conversations early so people aren’t caught off guard
- •She suggests society must rethink connections between work, meaning, and social life
- •She says AI will become a major social and political issue requiring multi-stakeholder involvement
AI adoption beyond Silicon Valley: demographics, global optimism, and “early in the game”
The discussion turns to why AI adoption varies outside elite tech bubbles. Daniela points to demographic skews in usage and notes that developing countries may be more optimistic than higher-income regions, arguing that the broader race of adoption and value-shaping has just begun.
- •She observes many people outside Silicon Valley aren’t yet comfortable or fluent with AI tools
- •She describes demographic differences in adoption (education, gender, race, wealth)
- •She notes uneven distribution globally and differing regional attitudes
- •She contrasts optimism in developing countries with anxiety in higher-income countries
- •She argues it’s still early, with significant opportunity to shape access, use, and embedded values
What we risk losing by delegating too much: disengagement vs “learning mode”
Daniela shares insights from a large qualitative survey about how AI affects people’s thinking and learning. She contrasts “turn your brain off” usage with a tutoring-like approach, describing a “learning mode” concept as a model for tools that help users grow rather than bypass effort.
- •She references a large qualitative survey of people’s AI use and feelings
- •She gives an example of AI enabling new capabilities, like building a website
- •She describes a risk: not engaging your brain because the tool makes it easy
- •She notes models can be wrong and warns against blind trust
- •She contrasts cheating-style use with a tutoring-style approach that helps users learn and get unstuck
Human skills that matter more: relationships, communication, and trust
Daniela explains that as AI takes on more task-oriented work, human-to-human skills become more valuable. She uses medicine as an example, arguing that while AI may improve diagnostics, it can’t replace examination, rapport, or the relational components that influence outcomes.
- •She predicts task-oriented job components will shift significantly due to AI tools
- •She argues humans prefer learning with and from other humans
- •She emphasizes communication and explanation as differentiators in workplaces
- •She expects AI to get strong at diagnostic-style reasoning
- •She highlights relational skills like bedside manner and trust-building as increasingly prized
Favorite AI use cases: leadership coaching, pattern-spotting, and parenting support
Daniela shares personal examples of how she uses Claude. She describes using it as a management coach for performance feedback and as a practical, empathetic aid for parenting challenges, particularly for overwhelmed parents navigating unreliable information online.
- •She describes using Claude to help write performance reviews and identify longer-term patterns
- •She notes Claude can surface feedback themes that are easy to miss day-to-day
- •She says Claude can also give her direct feedback based on reports’ input
- •She shares using Claude to support parenting, including potty training
- •She argues interactive, measured responses can be helpful amid overwhelming online misinformation
Lessons for builders: choose what you care about and unite business with doing good
Before Q&A, Daniela offers two takeaways for aspiring leaders. She stresses the importance of building something you deeply care about to endure hard periods, and she highlights a growing belief that business success and doing good can reinforce each other.
- •She urges pursuing work that inspires sustained commitment when things get hard
- •She frames passion as crucial for resilience through unpleasant stretches
- •She argues business and social impact need not be in tension
- •She praises founders who merge innovation with doing good
- •She suggests the desire to do good correlates with doing well
Student Q&A: AI bubble concerns, regulation tradeoffs, and privacy/medical use
Students ask about bubble risk, regulatory balance, and protecting privacy as AI touches sensitive data. Daniela discusses the capital intensity and compute bet underpinning the industry, calls for nuanced regulation developed with industry insight, and argues companies must protect data while users should apply skepticism—especially for medical guidance.
- •She flags capital expenditure and scarce compute as a valid industry-wide risk
- •She describes buying compute far ahead as a bet that future revenue sustains costs
- •She calls AI regulation nuanced and hopes the debate avoids politicized binaries
- •She favors regulators and companies collaborating: firms see abuse patterns; regulators enforce frameworks
- •On privacy, she says companies should be held accountable and notes Claude’s no-ads decision reflects sensitivity of AI conversations
- •She cautions against acting on medical outputs without professional confirmation
Rapid fire and closing: literature, working with her brother, books, naming, and trusting your instincts
In rapid fire, Daniela says she’d major in literature again and shares upsides and challenges of working with her brother. She mentions rediscovering a favorite book, jokes about earlier name ideas, and closes with advice that people often already know the right decision when facing a major choice.
- •She says she’d likely study literature again because she loves reading
- •She describes the benefit of deep sibling familiarity and candor at work
- •She notes the difficulty of separating personal and work relationships and the need to invest in both
- •She cites rereading a favorite book found in the office library
- •She recalls earlier, less successful company name ideas before “Anthropic”
- •She shares mentor advice: in big decisions, you often already know the right answer

Publisher timestamps

Get more out of YouTube videos.

Read along on your phone, search it beside the video, or ask your favorite AI app what was said.