Chris Olah on Pope Leo XIV’s encyclical “Magnifica humanitas”
Key Points
- Calls for external moral oversight of AI
- Three discernment priorities: poverty, flourishing, model nature
- Request for broad, long‑term societal collaboration
Summary
On May 25, 2026, Anthropic co‑founder Chris Olah presented remarks at the Vatican in response to Pope Leo XIV’s encyclical “Magnifica humanitas.” Olah argues AI is not purely a technical problem: incentives inside labs can conflict with the public good, model behavior and internal structure raise novel ethical questions, and society outside labs (including religious and civic institutions) must participate in ongoing discernment and oversight.
Key Points
- AI development is shaped by commercial, geopolitical, and personal incentives; independent external critics are essential to keep labs accountable.
- Olah identifies three priority questions for moral discernment:
- Duty to the global poor: how to share AI gains and protect those displaced by automation.
- Moral imagination for flourishing: what human, family, and societal flourishing look like in an AI era.
- Nature of models: interpretability work has revealed mysterious, introspection‑like internal states that warrant ethical and philosophical study.
- He calls for long‑term collaboration between engineers, faith communities, scholars, governments, and civil society to guide AI toward human flourishing.
Implications for engineers (practical takeaways)
- Build for auditability and external review; document incentives and deployment choices.
- Prioritize interpretability and safety research and make results available to independent reviewers.
- Engage interdisciplinary stakeholders early (policy makers, labor groups, global partners, ethicists) to plan equitable distribution and mitigation strategies.