Out Of AI℠ — All Things AI
Start here ↗
AI ETHICS TRAINING · WISDOM INTO PRACTICE

Better tools.
Better judgment.

Bring truth, human dignity and care into the way you use AI. Explore ethical wisdom across traditions and turn reflection into responsible action.

6 lessons · 12 scenarios · 34 optional reading guides

01 / A starting point

Start with the person.

Expanded edition · October 26, 2026

Explore how people have approached questions of truth, dignity, fairness, compassion and responsibility across diverse philosophies, wisdom traditions and ethical teachings. Use these perspectives to examine intentions, question assumptions, consider consequences and make thoughtful decisions about AI in business and everyday life.

START WITH A REAL DECISION

What are you asking AI to do?

Name the actual action, the good you intend and the evidence you have. Separate an attractive result from the means used to obtain it.

Outline the decision →
CONSIDER THE PEOPLE INVOLVED

Who benefits—and who bears the cost?

Consider privacy, vulnerability, freedom and the ability to challenge a mistake. Identify who can stop the process and repair harm.

Explore a situation →
How the sources are used and compared
  1. Distinguish authority. Read each source within its own tradition and historical context. Distinguish Scripture, Church teaching, theological arguments, philosophical reasoning, Confucian and Buddhist texts, company policies and our practical judgments. When this course claims agreement with Catholic moral teaching, the Catechism is the organizing reference. A shared ethical point does not make these different sources interchangeable. Not every papal statement or theologian’s opinion has the same authority.
  2. Examine the act, intention and circumstances. Benefits do not justify wrongful means. We do not reduce moral judgment to a numerical risk score. CCC 1750–1754 · The sources of morality · CCC 1755–1756 · Good acts and evil acts
  3. Read authors in context. Augustine’s account of love also requires correcting misinterpretation. Aquinas’s account of prudence joins inquiry and judgment to action. Alongside them, classical, Confucian and Buddhist texts are read as selected teachings with their own contexts. These are starting readings, not an exhaustive consultation of any author’s works. Where an individual theological opinion differs from later authoritative Church teaching, follow the Church’s teaching.
  4. Prefer truth to pleasing answers. Check exact passages, objections, limitations and empirical evidence. Identify missing information, resist rationalization and correct errors openly.
  5. Protect people. Respect confidentiality, conscience, dignity and the common good. A corporate discussion should not require a participant to profess religious belief or disclose private beliefs.
  6. Keep judgment accountable. Seek checkable reasons and observable conduct, not promises about a model’s hidden reasoning. Human review remains necessary.

The exercises are OutOfAI℠’s prudential applications. They do not carry ecclesiastical approval or provide professional certification. This selected-reading edition cannot guarantee exhaustive theological coverage or that an AI tool will follow the framework. Serious doctrinal questions merit qualified Catholic guidance; real regulated decisions also need the relevant professional review.

02 / Build the habit

Six lessons. Everyday decisions.

Go to the practice lab →
01Truth before persuasionCCC 2475–2487

Catholic foundation

Truth and a neighbor’s reputation must not be sacrificed to a persuasive result. Repair matters when false information has caused harm.

Reading in context

Augustine connects sound interpretation with love of God and neighbor while still requiring correction of errors. A flattering answer is not a substitute for a true one.

OutOfAI℠ practical application

List factual claims and their sources. Open each source. Remove unsupported claims; identify synthetic content where its omission would mislead. Record corrections and notify affected readers.

Starter prompt

Review this draft only against the supplied sources. Separate supported facts, interpretations and unsupported claims. Do not invent citations, testimonials or quotations. Flag misleading omissions and unjust reputational harm. Give a corrected draft and a short explanation that a human can verify. Identify any sources you could not access.

Read the sources: CCC 2475–2487 · Offenses against truth · Augustine · On Christian Doctrine I.36.40–41 · Amanda Askell / Anthropic · Claude’s Constitution · Michael Kearns · Responsible AI in the generative era

Practice this lesson →
02A person is more than a scoreCCC 1750–1756; 2426

Catholic foundation

A worthwhile goal does not excuse wrongful means. Examine what is actually being done to people, the intention and the concrete circumstances.

Reading in context

Leo XIV places human dignity and the common good at the center of reflection on AI. Applying this to a decision system requires attention to people who bear its mistakes.

OutOfAI℠ practical application

Describe the action in plain language. Identify affected people, relevant evidence and possible exclusions. Require meaningful human review and a way to contest an error before consequential use.

Starter prompt

Help prepare a human review of this proposed AI use. Describe the action chosen, intended purpose and circumstances separately. Identify affected people, uncertain evidence, unfair exclusions and ways to challenge mistakes. Do not rank real people, infer protected traits or make an eligibility decision. Suggest alternatives that preserve dignity without using wrongful means.

Read the sources: CCC 1750–1754 · The sources of morality · CCC 1755–1756 · Good acts and evil acts · CCC 2426–2428 · Economic life and work · Leo XIV · Magnifica Humanitas, §§46–50, 68–73 · Sandhini Agarwal · CoVal: Learning values-aware rubrics from the crowd

Practice this lesson →
03Permission before convenienceCCC 2488–2492

Catholic foundation

Being truthful does not require revealing every private fact. Respect the legitimate limits on who should receive information.

Reading in context

Leo XIV’s concern for human voices and faces invites careful judgment about imitation and identity. OutOfAI℠ applies this to recordings, synthetic voices and private uploads.

OutOfAI℠ practical application

Identify whose information is involved and the permitted purpose. Remove unnecessary details. Use approved tools and obtain the appropriate authorization before uploading, recording or imitating someone.

Starter prompt

Using a fictional or de-identified description, identify privacy, confidentiality and identity concerns in this AI task. List information that is unnecessary, permissions to verify, safer alternatives and questions for the responsible reviewer. Do not request secrets or personal records. Do not assume that possession of a file gives permission to upload or reuse it.

Read the sources: CCC 2488–2492 · Respect for the truth · Leo XIV · Preserving Human Voices and Faces

Practice this lesson →
04Let work serve peopleCCC 2426–2428

Catholic foundation

Economic activity should serve human needs. A person’s value cannot be reduced to a productivity measure.

Reading in context

Leo XIV’s emphasis on solidarity and subsidiarity invites participation by affected people. A practical policy should consider both shared benefit and the responsibility of those closest to the work.

OutOfAI℠ practical application

Consult the people doing the work. Assess workload, access, retraining and service quality alongside cost. Test a limited pilot and review its effects before expanding it.

Starter prompt

Compare this proposed automation with a limited pilot and a non-AI alternative. Consider effects on workers, customers, accessibility, skills and service quality as well as cost. Separate evidence from estimates. Identify who should be consulted and what to measure. Do not recommend individual employment decisions.

Read the sources: CCC 2426–2428 · Economic life and work · Leo XIV · Magnifica Humanitas, §§46–50, 68–73 · Natasha Crampton / Microsoft · Responsible AI in 2026 · Grace Yee · Reflecting on our five-year journey with our AI Ethics principles

Practice this lesson →
05Care beyond the screenCCC 2488–2492; Gospel application

Catholic foundation

Jesus’ teaching on love of neighbor and the Good Samaritan directs attention to the person in need, including those outside our usual circle.

Reading in context

Augustine’s account of ordered interpretation keeps love of God and neighbor in view. Leo XIV warns that simulated relationships can displace genuine encounter, especially for vulnerable people.

OutOfAI℠ practical application

Notice when a tool encourages dependency or substitutes for needed human contact. Make its automated nature clear, avoid manipulative engagement tactics and provide an accessible human route.

Starter prompt

Review this fictional AI interaction for manipulation, false intimacy, dependency and exclusion. Suggest a truthful disclosure of the tool’s role and a practical way to involve a trusted human. Respect privacy and avoid diagnosing anyone. Explain how the proposed response serves the person rather than merely increasing engagement.

Read the sources: Matthew 22:34–40 · Love of God and neighbor · Luke 10:25–37 · The Good Samaritan · Augustine · On Christian Doctrine I.36.40–41 · Leo XIV · Preserving Human Voices and Faces · CCC 2488–2492 · Respect for the truth

Practice this lesson →
06Someone must own the decisionCCC 1750–1756; 2487

Catholic foundation

People remain responsible for their choices. Good intentions and a model’s confidence do not settle whether an action is right.

Reading in context

Aquinas describes prudence as inquiry, judgment and the direction of action. OutOfAI℠ applies this by requiring a reviewer who can actually pause, correct or reject an AI-assisted action.

OutOfAI℠ practical application

Name an accountable role. Set approval limits, a stop condition and a review date. Keep a concise evidence record. If harm occurs, correct it, inform affected people appropriately and change the process.

Starter prompt

Prepare a review memo, not an approval, for this proposed action. State the facts, missing evidence, ethical objections, alternatives, approval authority, stop conditions and correction plan. Give concise, checkable reasons, not claims about a model’s hidden reasoning. Explicitly state what you cannot verify. Leave the final decision to an accountable human.

Read the sources: CCC 1750–1754 · The sources of morality · CCC 1755–1756 · Good acts and evil acts · CCC 2475–2487 · Offenses against truth · Aquinas · Summa Theologiae II–II, q.47, a.8 · Francesca Rossi / IBM · Human accountability for AI · Anca Dragan · Taking a responsible path to AGI

Practice this lesson →
03 / Practice lab

What would you do next?

0 of 12 scenarios reflected on this visit

Choose a situation, consider an action and read the reasoning. These fictional exercises invite discernment; they do not approve a real decision.

04 / Bring it to your work

Make the reasoning reviewable.

Use a fictional or de-identified example. Finish with a review memo—not an automated moral verdict.

Do not enter personal records, confidential business information or secrets. This worksheet does not send entries to an AI service or save them after reload. Existing site analytics may record visits. Download your draft if you want to keep it.

05 / Optional study & resources

Starting points for further exploration.

These 34 reading guides are optional starting points for anyone interested in further study. They introduce a range of philosophies, teachings and contemporary perspectives. Some may interest you; others may not. Explore what is useful to your questions, follow the original sources and form your own considered judgments. You can use the lessons, practice lab and decision worksheet without completing these readings.

Guides are presented in a shuffled order, without ranking or grouping by religion or philosophy. Filter by ethical topic or search for a name. Source identity and context remain visible within every guide.

Compare a particular teaching, not an entire tradition.
Dignity & compassion

Pope Leo XIV · Human dignity

Keep human encounter and responsibility centralOpen teaching & exercise +

Source tradition: Papal teaching

Teaching: In Magnifica Humanitas, Leo XIV connects the AI age with the common good, solidarity and subsidiarity. His communications message emphasizes responsibility, cooperation and education, and the importance of genuine human encounter.

Read in context: These are papal texts within Catholic teaching. The workflow below is OutOfAI℠’s application; it is not a papal endorsement of any AI tool.

OutOfAI℠ application

At work: Include people affected by an automated process in its review and preserve a meaningful route to human help.

Everyday life: Use AI assistance without allowing it to displace attention to people and responsibility for your own choices.

Starter prompt

Review this fictional AI proposal for human dignity, the common good, participation and responsibility. Identify who is affected, who should be consulted and how genuine human help remains available.

Leo XIV · Magnifica Humanitas, §§46–50, 68–73 · Leo XIV · Preserving Human Voices and Faces

Try a related scenario →
Care & relationships

St. Francis de Sales

Practice charity in ordinary dutiesOpen teaching & exercise +

Source tradition: Doctor of the Church

Teaching: Francis de Sales teaches a life devoted to God within ordinary responsibilities, with charity, courtesy and personal relationship rather than coercion.

Read in context: The papal catechesis draws on his spiritual writings. Gentleness does not require concealing truth or abandoning a just boundary.

OutOfAI℠ application

At work: Revise automated replies so they are clear, truthful and considerate rather than dismissive.

Everyday life: Use writing assistance without letting it turn disagreement into humiliation.

Starter prompt

Rewrite this fictional message with charity and clarity. Preserve the facts and necessary boundary, remove insults and pressure, and avoid false promises or misleading reassurance.

Benedict XVI · Francis de Sales · Francis — Totum amoris est (December 28, 2022)

Try a related scenario →
Privacy & reciprocity

Confucius · Reciprocity

Consider the other person’s positionOpen teaching & exercise +

Source tradition: Confucian ethical text

Teaching: Analects 15.24 presents reciprocity as avoiding imposing on others what one would not want imposed on oneself.

Read in context: This is a saying attributed to Confucius in the Analects. Its negative formulation overlaps with, but is not identical to, the Gospel’s positive command to act for others’ good.

OutOfAI℠ application

At work: Before feeding a customer’s confidential message into a tool, consider the permission and protection you would reasonably expect yourself.

Everyday life: Ask before using another person’s likeness or words in an AI-generated post.

Starter prompt

Examine this fictional AI use from each affected person’s position. Identify reasonable expectations, consent, privacy and possible burdens. Explain what respectful treatment requires, without assuming that my preferences are shared by everyone.

Analects 15.24 · Reciprocity · CCC 1786–1789 · Discernment and the Golden Rule · CCC 2488–2492 · Respect for the truth

Try a related scenario →
Responsibility & repair

Hans Jonas

Consider the people affected tomorrowOpen teaching & exercise +

Source tradition: Philosophical inquiry

Starting reading: The Imperative of Responsibility

University press overview; obtain the complete book through a library or publisher.

Teaching: Jonas examines how powerful technologies expand ethical responsibility toward future people and the environment. Immediate benefits do not settle questions about lasting effects.

Read in context: The book addresses the technological age broadly. The AI-agent example below is a present-day adaptation, not a claim that Jonas assessed current systems. His philosophical grounding is distinct from Church teaching.

OutOfAI℠ application

At work: Before deploying an agent that changes shared records, consider how errors could accumulate and affect people who cannot easily obtain a correction.

Everyday life: Consider how an AI-generated image or accusation could persist online and affect someone years after you post it.

Starter prompt

For a fictional AI rollout, identify immediate and longer-term effects, people not represented in the decision, reversible and hard-to-reverse actions, and uncertainties. Propose a proportionate pilot, monitoring and stop criteria. Label forecasts rather than presenting them as facts.

Hans Jonas · The Imperative of Responsibility · CCC 1750–1754 · The sources of morality · Aquinas · Summa Theologiae II–II, q.47, a.8

Try a related scenario →
Responsibility & repair

Aristotle

Practice builds characterOpen teaching & exercise +

Source tradition: Classical philosophy

Teaching: Nicomachean Ethics II.1 connects moral character with habitual action. VI.4–5 distinguishes productive skill from practical wisdom about human good.

Read in context: A capable tool does not make its use wise. This selected comparison is not acceptance of every Aristotelian view; classical philosophy remains subject to evaluation within Catholic teaching.

OutOfAI℠ application

At work: Make source checks and correction routine rather than optional steps under deadline pressure.

Everyday life: Use AI to support learning while retaining the habit of reasoning and checking your own work.

Starter prompt

Design a repeatable review habit for this fictional AI task: define the human good, identify the temptation to skip a check, specify the check, and explain how a human will notice and repair failure.

Aristotle · Nicomachean Ethics II.1 · Aristotle · Nicomachean Ethics VI.4–5

Try a related scenario →
Dignity & compassion

The Buddha · Compassion

Take another’s suffering seriouslyOpen teaching & exercise +

Source tradition: Buddhist ethical text

Teaching: Dhammapada 129–130 appeals to shared vulnerability and the value of life when urging restraint from killing and causing killing.

Read in context: The AI example extends the passage’s concern for suffering by analogy. It does not claim agreement on every Buddhist and Catholic teaching about nonviolence, the self or the ultimate purpose of life.

OutOfAI℠ application

At work: Review whether a cost-saving support system leaves distressed or vulnerable customers unable to reach help.

Everyday life: Consider the person who could be frightened or harmed by a synthetic image before sharing it.

Starter prompt

Review this fictional AI workflow for foreseeable suffering. Describe who may be vulnerable, what harm could occur, a less harmful alternative and a real route to human help. Keep uncertainty explicit.

Dhammapada 129–130 · Considering another’s suffering · Luke 10:25–37 · The Good Samaritan

Try a related scenario →
Responsibility & repair

Hannah Arendt

Keep judgment personally accountableOpen teaching & exercise +

Source tradition: Philosophical inquiry

Starting reading: Responsibility and Judgment

Publisher overview and excerpt; obtain the complete collection through a library or publisher.

Teaching: Arendt examines moral judgment and personal responsibility, including the limits of explanations that dissolve a person’s actions into a system or institution.

Read in context: This posthumous collection includes work shaped by the historical experience of totalitarianism. An ordinary AI error is not being equated with those historical crimes.

OutOfAI℠ application

At work: A manager approves an AI-generated rejection and later says the software made the decision. Identify who reviewed the evidence and who has authority to correct the outcome.

Everyday life: If an AI-written message harms someone, acknowledge your choice to send it and help repair the damage.

Starter prompt

Map the human decisions in this fictional AI-supported process: who set the goal, approved the data, reviewed the result and acted on it? Identify missing checks, correction authority and a record of reasons. Distinguish responsibility from blaming a person without evidence.

Hannah Arendt · Responsibility and Judgment · CCC 1750–1754 · The sources of morality

Try a related scenario →
Dignity & compassion

Jacques Maritain

Serve persons and the common goodOpen teaching & exercise +

Source tradition: Philosophical inquiry

Starting reading: The Person and the Common Good

Free university-hosted reading; choose “Read The Person and the Common Good” on the bookshelf.

Teaching: Maritain relates the dignity of persons to the good of the communities they share. His personalism rejects treating private advantage as the sole purpose of social life.

Read in context: This is a Thomistic philosophical argument. It is not itself a magisterial document, and its application to a particular workplace remains a matter for reasoned judgment.

OutOfAI℠ application

At work: An AI scheduling system reduces costs but creates unstable hours. Examine the effects on workers, customers and the organization before calling the result an improvement.

Everyday life: When choosing a household AI tool, consider the interests and consent of everyone whose information it will process.

Starter prompt

Evaluate a fictional AI scheduling change from the perspectives of workers, customers and the business. Identify benefits, burdens and whose voice is missing. Propose an alternative that serves shared needs while respecting each person. Do not reduce dignity to an efficiency score.

Jacques Maritain · The Person and the Common Good · CCC 2426–2428 · Economic life and work

Try a related scenario →
Habits & work

Alasdair MacIntyre

Build good habits, not just better scoresOpen teaching & exercise +

Source tradition: Philosophical inquiry

Starting reading: After Virtue

University press overview; obtain the complete book through a library or publisher.

Teaching: MacIntyre examines virtue and the practices through which people pursue worthwhile forms of excellence. His work invites questions about the character encouraged by organizational rewards.

Read in context: After Virtue is a philosophical argument, not a business checklist or Church document. The proposed audit of workplace incentives is OutOfAI℠’s application.

OutOfAI℠ application

At work: A support team is rewarded only for closing tickets quickly. Examine whether AI-generated replies encourage genuine service or merely improve the measured score.

Everyday life: Use a learning assistant to develop understanding, then practice unaided rather than treating completed answers as proof of skill.

Starter prompt

Audit the incentives in a fictional AI-assisted team. Identify the real purpose of the work, the qualities of excellent service and how current metrics could reward shortcuts. Propose measures and review habits that support both competence and integrity.

Alasdair MacIntyre · After Virtue · CCC 2426–2428 · Economic life and work

Try a related scenario →
Truth & evidence

Margaret Mitchell · Hugging Face

Make AI capabilities and limitations visibleOpen teaching & exercise +

Source tradition: Contemporary responsible-AI practice

Academic background: PhD in Computer Science, University of Aberdeen; master’s in Computational Linguistics, University of Washington. Documented in her Harvard Berkman Klein Center biography.

Scope of expertise: Computer science, language technology and ethics-informed machine learning. Relevant to evaluation, documentation and AI limitations; no philosophy or theology degree is claimed.

Degree evidence: Margaret Mitchell — role and background

Starting reading: Model Cards for Model Reporting

Open-access research paper

Teaching: Mitchell and her coauthors propose model cards: documentation that explains a model’s intended uses, evaluation conditions and performance across relevant groups. These reports help readers judge whether evidence supports a proposed use.

Read in context: Margaret Mitchell, Simone Wu, Andrew Zaldivar, Parker Barnes, Lucy Vasserman, Ben Hutchinson, Elena Spitzer, Inioluwa Deborah Raji and Timnit Gebru. October 5, 2018 preprint; revised January 14, 2019. This foundational research predates today’s generative-AI products. Documentation can expose gaps but does not itself prove safety, fairness or fitness for a particular deployment. Read the evaluation methods as well as the summary.

OutOfAI℠ application

At work: Before buying an AI support tool, request its model documentation. Identify the languages and customer groups tested, known failure modes and any proposed use that lacks evidence. Assign a person to resolve those gaps before launch.

Everyday life: Before relying on an AI answer, check what the provider says the system can and cannot do. Treat an untested capability as uncertain rather than inferring reliability from a confident response.

Starter prompt

For a fictional customer-support AI, draft a model-card review covering intended use, prohibited use, test conditions, performance across relevant user groups, limitations and missing evidence. Do not invent test results. End with three questions a buyer should resolve before approval.

Model Cards for Model Reporting · Margaret Mitchell — role and background · CCC 2475–2487 · Offenses against truth

Try a related scenario →
Care & relationships

Edith Stein

Attend to another person’s experienceOpen teaching & exercise +

Source tradition: Philosophical inquiry

Starting reading: On the Problem of Empathy

Publisher overview; obtain the complete book through a library or publisher.

Teaching: Stein investigates empathy and how another person’s experience becomes accessible to us. Her work invites careful attention to the person whose experience differs from our own.

Read in context: This philosophical work originated as her 1916 doctoral dissertation under Edmund Husserl. It is not a study of machine consciousness or evidence that an AI system feels empathy.

OutOfAI℠ application

At work: A chatbot closes a customer complaint after detecting a polite reply. Ask whether the customer was actually helped and provide a human route for unresolved needs.

Everyday life: When a friend shares a difficulty, use an AI draft only as preparation. Listen to the person rather than assuming the draft captures their experience.

Starter prompt

Using a fictional customer complaint, identify what the customer actually said, what we are assuming about their experience, and what respectful follow-up could clarify. Propose a human escalation path. Do not infer private feelings as facts or claim that AI experiences empathy.

Edith Stein · On the Problem of Empathy · Luke 10:25–37 · The Good Samaritan

Try a related scenario →
Responsibility & repair

St. Thomas Aquinas

Turn prudent judgment into actionOpen teaching & exercise +

Source tradition: Doctor of the Church

Teaching: Summa Theologiae II–II, q.47, a.8 treats command—the application of counsel and judgment to action—as prudence’s principal act.

Read in context: Read Aquinas’s response and replies alongside the objections. Prudence works within moral truth; it is not simply finding an efficient way to get what one wants.

OutOfAI℠ application

At work: Give the human reviewer genuine authority to stop an agent, not merely a title.

Everyday life: Move from identifying a false post to making and communicating a correction.

Starter prompt

For this fictional AI use, list the proposed act, intention and circumstances. Set out relevant objections, alternatives, the person’s review role and the concrete action needed. Do not substitute an efficiency score for moral judgment.

Aquinas · Summa Theologiae II–II, q.47, a.8 · CCC 1750–1754 · The sources of morality · CCC 1755–1756 · Good acts and evil acts

Try a related scenario →
Responsibility & repair

Luciano Floridi · Yale Digital Ethics Center

Turn ethical principles into practical checksOpen teaching & exercise +

Source tradition: Contemporary responsible-AI practice

Academic background: MPhil and PhD in Philosophy, University of Warwick, documented by the Oxford Internet Institute. Yale’s current profile also confirms his Warwick degrees.

Scope of expertise: Philosophy of information, digital ethics and AI ethics. Relevant to ethical reasoning and governance; no theology degree or engineering qualification is claimed.

Degree evidence: Oxford Internet Institute — Luciano Floridi biography

Starting reading: From What to How: An Initial Review of Publicly Available AI Ethics Tools, Methods and Research to Translate Principles into Practices

Open-access research paper

Teaching: Morley, Floridi and their coauthors examine the gap between stating ethical principles and applying them. Their review organizes tools, methods and research to help developers address ethics at different stages of machine-learning development.

Read in context: Jessica Morley, Luciano Floridi, Libby Kinsey and Anat Elhalal. May 15, 2019 preprint; revised September 13, 2019. This is a 2019 research review, not an up-to-date buying list or a certification scheme. Its useful starting question is how a principle changes a concrete decision. Floridi is included as an academic digital-ethics leader, not as an employee of an AI company.

OutOfAI℠ application

At work: For an AI procurement workflow, translate each stated principle into a question, evidence requirement, decision owner and review point. Revisit the decision when the model, data or intended use changes.

Everyday life: Turn “use AI responsibly” into a short routine: identify the purpose, consider who could be affected, check the output and decide when a person should take over.

Starter prompt

For a fictional AI purchasing decision, turn truthfulness, dignity and accountability into review steps. For each, specify a question, evidence to request, a responsible person, a stop condition and a later review trigger. Mark unknowns and avoid treating a completed checklist as proof that every ethical issue is resolved.

From What to How: An Initial Review of Publicly Available AI Ethics Tools, Methods and Research to Translate Principles into Practices · Luciano Floridi — role and background · CCC 1750–1754 · The sources of morality · Oxford Internet Institute — Luciano Floridi biography

Try a related scenario →
Truth & evidence

St. Augustine of Hippo

Let truth serve charityOpen teaching & exercise +

Source tradition: Doctor of the Church

Teaching: On Christian Doctrine I.36 connects interpretation with love of God and neighbor, while still requiring mistaken interpretation to be corrected.

Read in context: Charitable intention does not make a false interpretation true. This reading is a starting point within Augustine’s wider theological work, not a complete account.

OutOfAI℠ application

At work: Correct an AI-written statement that misrepresents a colleague even when it would make your project look stronger.

Everyday life: Read the whole passage before circulating an AI-generated religious quotation.

Starter prompt

Review this fictional summary for unsupported interpretations and harm to another person’s reputation. Separate the source’s actual meaning from my assumptions; suggest a truthful and charitable correction.

Augustine · On Christian Doctrine I.36.40–41 · CCC 2475–2487 · Offenses against truth

Try a related scenario →
Responsibility & repair

Contemporary responsible-AI practice

Make oversight observableOpen teaching & exercise +

Source tradition: Contemporary responsible-AI practice

Teaching: Selected perspectives from Amanda Askell, Natasha Crampton and Francesca Rossi emphasize honesty, governance as risks change, and human accountability. These are complementary points from separate sources, not a joint statement.

Read in context: These professional perspectives are not religious doctrine or evidence that a company’s products always meet its stated principles. OutOfAI℠ compares the particular points described, without claiming endorsement by the authors.

OutOfAI℠ application

At work: Define the reviewer’s authority, monitor changing permissions and record how mistakes are corrected.

Everyday life: Check an assistant’s uncertainty and sources before relying on an answer, and remain responsible for what you share.

Starter prompt

For this fictional AI process, specify evidence checks, human approval limits, a change that triggers renewed review, a stop mechanism and a correction record. Distinguish stated principles from controls that have actually been tested.

Amanda Askell / Anthropic · Claude’s Constitution · Natasha Crampton / Microsoft · Responsible AI in 2026 · Francesca Rossi / IBM · Human accountability for AI

Try a related scenario →
Dignity & compassion

Immanuel Kant

Treat people as more than a meansOpen teaching & exercise +

Source tradition: Philosophical inquiry

Starting reading: Groundwork of the Metaphysics of Morals · Section II

Free full text, published in this translation as Fundamental Principles of the Metaphysic of Morals.

Teaching: Kant asks whether the rule behind an action can be universalized and whether persons are respected as ends, rather than treated merely as instruments for another’s purpose.

Read in context: This guide selects two moral tests from Kant’s framework. His account of autonomy and the foundations of morality is not identical to Catholic natural-law reasoning.

OutOfAI℠ application

At work: An AI sales assistant exploits a customer’s anxiety to push an unnecessary purchase. Review whether the interaction permits informed, unmanipulated choice.

Everyday life: Before using AI to influence a friend, ask whether you could honestly disclose the method and still respect their freedom to disagree.

Starter prompt

Examine this fictional AI persuasion tactic. State its underlying rule, consider applying that rule consistently and ask whether affected people are treated merely as tools. Identify manipulative elements and propose a transparent alternative. Note the limits of this philosophical test.

Immanuel Kant · Groundwork of the Metaphysics of Morals · Section II · CCC 1786–1789 · Discernment and the Golden Rule · CCC 1755–1756 · Good acts and evil acts

Try a related scenario →
Truth & evidence

Michael Kearns · Amazon

Truth and fairness in generated contentOpen teaching & exercise +

Source tradition: Contemporary responsible-AI practice

Academic background: Undergraduate degree in mathematics and computer science, University of California, Berkeley; Ph.D. in Computer Science, Harvard University, documented by the University of Pennsylvania.

Scope of expertise: Mathematics and computer science. Relevant to machine learning, algorithmic fairness and evaluating technical claims; the Ph.D. subject is computer science, not academic philosophy.

Degree evidence: University of Pennsylvania · Michael Kearns’s mathematics and computer science education

Teaching: Kearns examines fairness, harmful content, fabricated information, intellectual property and other challenges of generative AI. He emphasizes that defining and evaluating these problems can depend on the specific use.

Read in context: Michael Kearns. May 3, 2023. This foundational 2023 essay is not a current product benchmark. Kearns is an Amazon Scholar and academic researcher, not presented here as Amazon’s chief ethics officer.

OutOfAI℠ application

At work: An assistant drafts a proposal containing customer testimonials and citations. Check that every quotation, source and performance claim is real and authorized before publication.

Everyday life: Open the original source before reposting an AI-generated claim. Correct a false statement already shared rather than quietly replacing it.

Starter prompt

Review this fictional AI proposal. Separate verifiable claims, opinions and unsupported assertions; flag fabricated citations, unfair assumptions and possible copying. List evidence needed, revisions and a correction plan. Never invent a source to make the draft look complete.

Michael Kearns · Responsible AI in the generative era · Michael Kearns · Role and background · CCC 2475–2487 · Offenses against truth · University of Pennsylvania · Michael Kearns’s mathematics and computer science education

Try a related scenario →
Truth & evidence

Blaise Pascal

Question persuasive appearancesOpen teaching & exercise +

Source tradition: Philosophical inquiry

Starting reading: Pensées · Selected reflections on imagination and self-knowledge

Free text; fragment numbering varies between editions.

Teaching: Pascal examines the ways imagination, appearances and self-interest can distort judgment. Intellectual skill does not remove the need for humility about our own reasoning.

Read in context: The Pensées are fragments assembled after his death, not a finished systematic textbook. His seventeenth-century religious controversies require context; this guide selects an insight about self-deception.

OutOfAI℠ application

At work: A fluent AI report confirms a preferred investment in a project. Ask for independent evidence, uncertainties and a serious alternative interpretation before accepting it.

Everyday life: Notice whether you trust an answer because it is well supported or because it flatters a belief you already hold.

Starter prompt

Review this fictional AI recommendation for unsupported confidence, persuasive presentation and confirmation bias. Separate evidence from assumptions and list an objection that could change the decision. Do not confuse eloquence with accuracy or skepticism with proof that a claim is false.

Blaise Pascal · Pensées · Selected reflections on imagination and self-knowledge · CCC 2475–2487 · Offenses against truth

Try a related scenario →
Truth & evidence

Plato

Look beyond convincing appearancesOpen teaching & exercise +

Source tradition: Classical philosophy

Teaching: The cave in Republic VII depicts people mistaking shadows for reality. Education involves a demanding reorientation toward understanding and the good.

Read in context: Connecting the cave to synthetic media is OutOfAI℠’s analogy. We do not identify Plato’s Form of the Good with Christian revelation or adopt his political system wholesale.

OutOfAI℠ application

At work: Trace an impressive AI presentation back to original data instead of judging its visual polish.

Everyday life: Verify an image or video through reliable independent evidence before treating it as a record of an event.

Starter prompt

For this fictional presentation, distinguish appearance, factual claim and supporting evidence. Identify what an audience could mistakenly infer and propose a transparent correction. Mark anything you cannot verify.

Plato · Republic VII, 514a–521b

Try a related scenario →
Responsibility & repair

Anca Dragan · Google DeepMind

Prudence before greater autonomyOpen teaching & exercise +

Source tradition: Contemporary responsible-AI practice

Academic background: B.Sc. in Computer Science, Jacobs University Bremen; Ph.D. in Robotics, Carnegie Mellon University, documented by UC Berkeley’s Center for Human-Compatible AI.

Scope of expertise: Computer science and robotics. Relevant to AI alignment, human–robot interaction and technical safety; technical expertise does not confer authority on Catholic doctrine.

Degree evidence: UC Berkeley CHAI · Anca Dragan’s computer science and robotics degrees

Teaching: The authors argue that increasingly capable AI requires proactive assessment of risks, preparation and collaboration. Potential benefits do not remove the need to take serious possibilities of harm into account.

Read in context: Anca Dragan, Rohin Shah, Four Flynn and Shane Legg (coauthors). April 2, 2025. This article concerns advanced AI safety and future capabilities. It does not establish a certain AGI timeline. Dragan’s role is safety and alignment research; the enterprise exercise below is OutOfAI℠’s adaptation.

OutOfAI℠ application

At work: An AI agent will move from drafting purchase orders to sending them. Reassess permissions, spending limits, monitoring, approval thresholds and a tested stop mechanism before expansion.

Everyday life: Keep review enabled before an assistant sends messages, makes purchases or changes shared information on your behalf.

Starter prompt

For a fictional purchasing agent, identify its proposed new powers, plausible misuse and failure paths, affected people, approval limits, monitoring, stop conditions and recovery steps. State what evidence is required before a human authorizes a limited trial.

Anca Dragan · Taking a responsible path to AGI · Anca Dragan · Role and background · Aquinas · Summa Theologiae II–II, q.47, a.8 · UC Berkeley CHAI · Anca Dragan’s computer science and robotics degrees

Try a related scenario →
Habits & work

Grace Yee · Adobe

Turn principles into review practicesOpen teaching & exercise +

Source tradition: Contemporary responsible-AI practice

Academic background: Bachelor of Science in Engineering, University of Michigan, Ann Arbor, listed in the American Marketing Association’s Adobe webinar biography.

Scope of expertise: Engineering background. Relevant to product development and implementing review processes. The source does not specify the engineering major; no philosophy or theology degree is claimed.

Degree evidence: American Marketing Association · Grace Yee’s engineering education

Teaching: Yee explains Adobe’s emphasis on accountability, responsibility and transparency, and how ethical review, feedback and practical guidance can be integrated into product development.

Read in context: Grace Yee. October 14, 2024. This is an account of Adobe’s own program, not an independent audit. A policy becomes useful in a workflow only when people can identify the review, decision owner and correction process.

OutOfAI℠ application

At work: Before launching an AI-generated marketing campaign, review accuracy, permissions, accessibility and likely effects on the audience. Record who can pause publication and correct a problem.

Everyday life: Review an AI-created image or message for misleading implications and another person’s rights before sharing it.

Starter prompt

Create a pre-publication review for a fictional AI marketing campaign. Include accuracy, permissions, accessibility, transparency, foreseeable harms, a named decision owner and a correction route. Give one example that must be revised before approval.

Grace Yee · Reflecting on our five-year journey with our AI Ethics principles · Grace Yee · Role and background · CCC 2426–2428 · Economic life and work · American Marketing Association · Grace Yee’s engineering education

Try a related scenario →
Responsibility & repair

Marcus Aurelius

Use power with restraint and self-examinationOpen teaching & exercise +

Source tradition: Classical philosophy

Starting reading: Meditations · Selected reflections on conduct and judgment

Free full text; section numbering and wording vary between translations.

Teaching: Marcus’s personal reflections repeatedly return to examining his own conduct, responding with restraint and acting with regard for other people.

Read in context: These are a Roman emperor’s Stoic reflections, not a comprehensive account of his rule or a guarantee that he always lived by his ideals. Stoic cosmology is not Christian doctrine.

OutOfAI℠ application

At work: After an AI rollout fails, a leader resists defensiveness, hears affected staff and makes corrections before attempting to protect the project’s reputation.

Everyday life: Pause before forwarding an angry AI-written reply; check whether it is truthful, proportionate and likely to help resolve the problem.

Starter prompt

For a fictional AI incident, draft a leader’s review that acknowledges known facts, hears affected people, avoids retaliation and assigns corrective actions. Distinguish composure from indifference and accountability from public image management.

Marcus Aurelius · Meditations · Selected reflections on conduct and judgment · Aquinas · Summa Theologiae II–II, q.47, a.8

Try a related scenario →
Dignity & compassion

St. John Chrysostom

Put the person before the systemOpen teaching & exercise +

Source tradition: Doctor of the Church

Teaching: Benedict XVI’s account of Chrysostom’s social teaching emphasizes the person’s dignity, including the poor, and responsibility beyond occasional assistance.

Read in context: This is a selected theme in a much larger body of preaching. It does not endorse every historical argument or expression in his works.

OutOfAI℠ application

At work: Check whether an automated support process leaves vulnerable customers unable to obtain help.

Everyday life: Consider the real person affected by an automated decision, not just a score or category.

Starter prompt

Audit this fictional service workflow from the perspective of a person with limited money, literacy or access. Identify obstacles, a working human escalation route and a way to verify that help reaches them.

Benedict XVI · John Chrysostom · Catechism 2443–2449 — Love for the poor

Try a related scenario →
Responsibility & repair

Confucius · Correcting faults

Treat correction as part of responsibilityOpen teaching & exercise +

Source tradition: Confucian ethical text

Teaching: Analects 15.30 identifies failing to correct a fault as itself a fault.

Read in context: The passage encourages moral self-correction. A modern incident procedure is OutOfAI℠’s application, not a process specified in the ancient text.

OutOfAI℠ application

At work: When an AI summary misstates a customer commitment, correct the record and address reliance on the mistake.

Everyday life: Tell the original audience when an AI-generated claim you shared proves false.

Starter prompt

Prepare a correction plan for this fictional AI error. State what is known, who received the error, how to correct the record, what harm may need repair and what safeguard should prevent repetition.

Analects 15.30 · Correcting faults · CCC 2475–2487 · Offenses against truth

Try a related scenario →
Truth & evidence

Socrates

Question confident answersOpen teaching & exercise +

Source tradition: Classical philosophy

Teaching: In Plato’s Apology, Socrates distinguishes recognizing one’s lack of knowledge from claiming knowledge one does not possess.

Read in context: We encounter Socrates here through Plato’s portrayal. This is a philosophical starting point for inquiry, not Christian revelation or a transcript of Socrates teaching about AI.

OutOfAI℠ application

At work: Before approving a confident AI report, ask which claims lack evidence and what would disprove the conclusion.

Everyday life: Ask what you actually know before sharing an AI-generated explanation.

Starter prompt

Examine this non-confidential claim. Define its key terms, separate evidence from assumptions, state what remains unknown, and suggest a check that could show it is mistaken. Do not invent sources or agree merely to please me.

Plato · Apology 21b–e

Try a related scenario →
Responsibility & repair

St. John Henry Newman

Form conscience; do not outsource itOpen teaching & exercise +

Source tradition: Doctor of the Church

Teaching: In chapter 5 of the Letter to the Duke of Norfolk, Newman distinguishes conscience as moral obligation before God from personal preference or self-will.

Read in context: Read his argument in context. Conscience requires formation and does not turn an AI recommendation or personal desire into moral authority. His Doctor declaration is confirmed by the Vatican source.

OutOfAI℠ application

At work: Require an accountable person to examine a troubling automated decision rather than hiding behind policy or software.

Everyday life: Seek sound reasons and qualified guidance when an AI answer conflicts with a serious moral duty.

Starter prompt

Help organize this fictional moral question for human review. Identify relevant duties, evidence, affected people and unresolved objections. Distinguish moral reasoning from preference; do not claim to replace conscience or give final moral approval.

John Henry Newman · Letter to the Duke of Norfolk, chapter 5 · Saint John Henry Newman · Doctor of the Church

Try a related scenario →
Truth & evidence

The Buddha · Thoughtful speech

Speak truth with care and timingOpen teaching & exercise +

Source tradition: Buddhist ethical text

Teaching: In Majjhima Nikaya 58, speech is examined for truth and benefit, with attention to the right time; an unwelcome truth may still need to be spoken.

Read in context: This is a discourse attributed to the Buddha in the Pali canon. It is not permission to tell a useful lie or to suppress necessary facts merely because they are uncomfortable.

OutOfAI℠ application

At work: Check whether an AI-drafted announcement is accurate, useful to its audience and timed responsibly.

Everyday life: Pause before sending an AI-written reply that is technically true but needlessly exposes someone’s private information.

Starter prompt

Assess this fictional message for truth, legitimate purpose, privacy and timing. Flag unsupported claims and unnecessary disclosure. Preserve any necessary unwelcome truth and suggest a respectful revision.

Majjhima Nikaya 58 · To Prince Abhaya · CCC 2475–2487 · Offenses against truth · CCC 2488–2492 · Respect for the truth

Try a related scenario →
Dignity & compassion

Sandhini Agarwal · OpenAI

Make value disagreements explicitOpen teaching & exercise +

Source tradition: Contemporary responsible-AI practice

Academic background: Bachelor of Arts and Science in Philosophy and Symbolic Systems, Stanford University; the university’s degree record lists conferral on January 9, 2020.

Scope of expertise: Philosophy and Symbolic Systems. Relevant to values, reasoning and human–AI interaction. This is an undergraduate qualification; no doctorate or theology degree is claimed.

Degree evidence: Stanford · Sandhini Agarwal’s conferred B.A.S. degree

Teaching: CoVal explores how people’s values affect their evaluations of AI responses. Its explicit rubrics make reasons for preference more visible and allow differences in model behavior and human judgments to be examined.

Read in context: Zoë Hitzig, Mitchell Gordon, Tyna Eloundou, Adam Kalai and Sandhini Agarwal (coauthors). January 14, 2026. The authors present an experimental method and dataset, not a universal moral standard. Surveyed preferences do not represent everyone or determine what is morally right. The study does not claim its rubrics express OpenAI’s own views.

OutOfAI℠ application

At work: Compare two AI-written customer responses, one blunt and one reassuring. Evaluate accuracy, empathy, clarity and avoidance of manipulation; explain disagreements without asking colleagues to disclose private religious beliefs.

Everyday life: When an AI response feels reassuring, ask whether it is also truthful and respectful of the people involved. Consider a reasonable objection before acting.

Starter prompt

Compare two fictional customer-service replies using explicit criteria for truth, dignity, empathy, clarity and avoidance of manipulation. Identify whose interests are affected, where reasonable preferences differ and which moral limits cannot be overridden by popularity. Explain a reviewable decision without requesting hidden chain-of-thought.

Sandhini Agarwal · CoVal: Learning values-aware rubrics from the crowd · Sandhini Agarwal · Role and background · CCC 1786–1789 · Discernment and the Golden Rule · Stanford · Sandhini Agarwal’s conferred B.A.S. degree

Try a related scenario →
Care & relationships

The Buddha · Responding to hostility

Do not multiply hatredOpen teaching & exercise +

Source tradition: Buddhist ethical text

Teaching: Dhammapada 5 teaches that hatred is not overcome through further hatred.

Read in context: This teaching is transmitted in the Buddhist canon. Responding without hatred does not require ignoring abuse, abandoning justice or surrendering appropriate protection.

OutOfAI℠ application

At work: Prevent an automated response to a hostile review from humiliating or threatening the reviewer.

Everyday life: Do not use an AI tool to amplify a feud or produce retaliatory insults.

Starter prompt

Rewrite this fictional hostile exchange to preserve verified facts and a necessary boundary while removing revenge, humiliation and threats. Include an appropriate next step for resolving the issue or seeking help.

Dhammapada 5 · Responding without hatred · CCC 2302–2305 · Safeguarding peace

Try a related scenario →
Truth & evidence

Cicero

Test apparent advantage against integrityOpen teaching & exercise +

Source tradition: Classical philosophy

Starting reading: On Duties (De Officiis) · Book III

Free full text; Walter Miller translation.

Teaching: Cicero examines apparent conflicts between honorable conduct and personal advantage. A profitable action still requires scrutiny of its honesty and duties toward others.

Read in context: This work belongs to the Roman world and draws on Greek ethical debate. Its social assumptions and historical examples are not all adopted here; the focus is integrity in decision-making.

OutOfAI℠ application

At work: An AI tool invents customer testimonials that improve conversion. Reject fabrication and seek truthful ways to communicate value.

Everyday life: Do not submit an AI-enhanced résumé containing abilities or experiences you cannot honestly substantiate.

Starter prompt

Assess a fictional AI marketing proposal that promises higher sales. Identify the claimed advantage, duties to customers, misleading statements and concealed assumptions. Propose a truthful alternative. Explain why a desirable outcome does not excuse dishonest means.

Cicero · On Duties (De Officiis) · Book III · CCC 1755–1756 · Good acts and evil acts · CCC 2475–2487 · Offenses against truth

Try a related scenario →
Responsibility & repair

Epictetus

Separate control from wishful thinkingOpen teaching & exercise +

Source tradition: Classical philosophy

Starting reading: Enchiridion · Sections 1 and 5

Free full text; Thomas Wentworth Higginson translation.

Teaching: Epictetus distinguishes what depends on our choices from what does not, and asks us to examine the judgments through which we respond to events.

Read in context: The handbook preserves teachings recorded by Arrian. Selected Stoic exercises are useful for reflection; they should not be used to blame people for suffering or excuse preventable harm.

OutOfAI℠ application

At work: A supplier’s AI service produces an error. Your team cannot guarantee the model’s behavior, but can change review rules, permissions and the response to affected customers.

Everyday life: Before sharing an alarming AI answer, pause and check the evidence. Distinguish the answer itself from your immediate reaction to it.

Starter prompt

For this fictional AI failure, separate what we can directly decide, what we can influence and what is outside our control. Identify duties to affected people and one corrective action. Do not use limited control as an excuse to abandon responsibility.

Epictetus · Enchiridion · Sections 1 and 5 · Aquinas · Summa Theologiae II–II, q.47, a.8

Try a related scenario →
Care & relationships

St. Basil the Great

Discern what is worth adoptingOpen teaching & exercise +

Source tradition: Doctor of the Church

Teaching: Basil recommends drawing from classical literature what accords with truth and virtue. His social teaching also treats possessions as a responsibility toward people in need.

Read in context: Benedict XVI’s catechesis explains these themes. Selective learning from pagan authors is neither wholesale rejection nor uncritical acceptance.

OutOfAI℠ application

At work: Evaluate whether automation benefits customers and workers who have the least ability to absorb its failures.

Everyday life: Keep useful insights from an AI answer while rejecting falsehood or harmful advice.

Starter prompt

Review this fictional AI proposal for what is true and useful, what must be rejected, and who might be excluded from its benefits. Suggest a concrete way to serve people with fewer resources.

Benedict XVI · Basil the Great

Try a related scenario →
Care & relationships

Jesus’ teachings · Love of neighbor

Make care concreteOpen teaching & exercise +

Source tradition: Christian Scripture

Teaching: Matthew 22:34–40 joins love of God with love of neighbor. The Good Samaritan in Luke 10:25–37 presents mercy that crosses social boundaries and responds to another’s need.

Read in context: These passages are read as Christian Scripture, not merely philosophical maxims. The AI example is a contemporary application rather than the original subject of the passages.

OutOfAI℠ application

At work: Look for a person left without assistance by an efficient automated process and provide a practical remedy.

Everyday life: Consider whether a generated message serves another person’s real need or simply your own convenience.

Starter prompt

Examine this fictional situation from the position of the person being overlooked. Name their need, one concrete act of help, its practical limits and who should take responsibility for following through.

Matthew 22:34–40 · Love of God and neighbor · Luke 10:25–37 · The Good Samaritan

Try a related scenario →
Truth & evidence

Confucius · Intellectual humility

Name the limits of your knowledgeOpen teaching & exercise +

Source tradition: Confucian ethical text

Teaching: Analects 2.17 distinguishes acknowledging what one knows from acknowledging what one does not know.

Read in context: This reading concerns honesty about knowledge. It does not imply that every question is unknowable or that evidence and expertise have no authority.

OutOfAI℠ application

At work: Require a report to separate verified findings from guesses before it is used in a business decision.

Everyday life: Do not pass along a chatbot’s confident answer as something you personally verified.

Starter prompt

Review this fictional answer. Separate supported facts, reasonable inferences and unknowns. Identify the evidence needed to resolve each important uncertainty; do not invent confidence percentages or sources.

Analects 2.17 · Knowing and not knowing · CCC 2475–2487 · Offenses against truth

Try a related scenario →
06 / Responsible-AI reading

Learn from responsible AI practice.

Selected ideas from AI ethicists and responsible-AI leaders can support these exercises. The connections below are OutOfAI℠’s comparisons, not claims that these contributors or their organizations endorse Catholic doctrine or this course.

How contributor qualifications are checked

A title alone is not an academic qualification. For these six added contributors, OutOfAI℠ checks documented degrees relevant to the subject matter, the author’s research or professional work, and the scope of the cited reading. Verified backgrounds cover engineering, mathematics, computer science, robotics, computational linguistics and philosophy. A Ph.D. is not necessarily a degree in philosophy. Academic credentials do not establish theological authority or make an argument correct. No theology degree was verified for these six contributors. Degree checks use public university records and published biographies, not private transcripts; checked October 26, 2026.

ANTHROPIC

Amanda Askell

Primary author of Claude’s Constitution. Its emphasis on honesty, uncertainty and avoiding empty validation offers a practical connection to our truthfulness lesson.

Apply it: ask for evidence and objections, not agreement.

Amanda Askell / Anthropic · Claude’s Constitution

MICROSOFT

Natasha Crampton

Chief Responsible AI Officer. Her account of adaptive governance and technical risk management connects with the need for review as systems and uses change.

Apply it: define a review point when permissions or consequences expand.

Natasha Crampton / Microsoft · Responsible AI in 2026

IBM

Francesca Rossi

AI ethics and governance leader. Her account of human accountability connects with the responsibility of people who deploy and rely on AI.

Apply it: name a reviewer with authority to stop and correct the action.

Francesca Rossi / IBM · Human accountability for AI

Adobe

Grace Yee

Senior Director of Ethical Innovation (AI Ethics and Accessibility)

Academic background: Bachelor of Science in Engineering, University of Michigan, Ann Arbor, listed in the American Marketing Association’s Adobe webinar biography.

Scope: Engineering background. Relevant to product development and implementing review processes. The source does not specify the engineering major; no philosophy or theology degree is claimed.

Degree evidence: American Marketing Association · Grace Yee’s engineering education

Turn principles into review practices

Reflecting on our five-year journey with our AI Ethics principles

Grace Yee · October 14, 2024

Open reading & exercise →
Google DeepMind

Anca Dragan

Head of AI Safety and Alignment; UC Berkeley professor on leave

Academic background: B.Sc. in Computer Science, Jacobs University Bremen; Ph.D. in Robotics, Carnegie Mellon University, documented by UC Berkeley’s Center for Human-Compatible AI.

Scope: Computer science and robotics. Relevant to AI alignment, human–robot interaction and technical safety; technical expertise does not confer authority on Catholic doctrine.

Degree evidence: UC Berkeley CHAI · Anca Dragan’s computer science and robotics degrees

Prudence before greater autonomy

Taking a responsible path to AGI

Anca Dragan, Rohin Shah, Four Flynn and Shane Legg (coauthors) · April 2, 2025

Open reading & exercise →
Amazon

Michael Kearns

Amazon Scholar and University of Pennsylvania professor

Academic background: Undergraduate degree in mathematics and computer science, University of California, Berkeley; Ph.D. in Computer Science, Harvard University, documented by the University of Pennsylvania.

Scope: Mathematics and computer science. Relevant to machine learning, algorithmic fairness and evaluating technical claims; the Ph.D. subject is computer science, not academic philosophy.

Degree evidence: University of Pennsylvania · Michael Kearns’s mathematics and computer science education

Truth and fairness in generated content

Responsible AI in the generative era

Michael Kearns · May 3, 2023

Open reading & exercise →
OpenAI

Sandhini Agarwal

AI policy researcher; role described in Caltech’s March 2026 speaker biography

Academic background: Bachelor of Arts and Science in Philosophy and Symbolic Systems, Stanford University; the university’s degree record lists conferral on January 9, 2020.

Scope: Philosophy and Symbolic Systems. Relevant to values, reasoning and human–AI interaction. This is an undergraduate qualification; no doctorate or theology degree is claimed.

Degree evidence: Stanford · Sandhini Agarwal’s conferred B.A.S. degree

Make value disagreements explicit

CoVal: Learning values-aware rubrics from the crowd

Zoë Hitzig, Mitchell Gordon, Tyna Eloundou, Adam Kalai and Sandhini Agarwal (coauthors) · January 14, 2026

Open reading & exercise →
Hugging Face

Margaret Mitchell

Researcher and Chief Ethics Scientist

Academic background: PhD in Computer Science, University of Aberdeen; master’s in Computational Linguistics, University of Washington. Documented in her Harvard Berkman Klein Center biography.

Scope: Computer science, language technology and ethics-informed machine learning. Relevant to evaluation, documentation and AI limitations; no philosophy or theology degree is claimed.

Degree evidence: Margaret Mitchell — role and background

Make AI capabilities and limitations visible

Model Cards for Model Reporting

Margaret Mitchell, Simone Wu, Andrew Zaldivar, Parker Barnes, Lucy Vasserman, Ben Hutchinson, Elena Spitzer, Inioluwa Deborah Raji and Timnit Gebru · October 5, 2018 preprint; revised January 14, 2019

Open reading & exercise →
Yale Digital Ethics Center

Luciano Floridi

Founding Director; John K. Castle Professor in the Practice of Cognitive Science

Academic background: MPhil and PhD in Philosophy, University of Warwick, documented by the Oxford Internet Institute. Yale’s current profile also confirms his Warwick degrees.

Scope: Philosophy of information, digital ethics and AI ethics. Relevant to ethical reasoning and governance; no theology degree or engineering qualification is claimed.

Degree evidence: Oxford Internet Institute — Luciano Floridi biography

Turn ethical principles into practical checks

From What to How: An Initial Review of Publicly Available AI Ethics Tools, Methods and Research to Translate Principles into Practices

Jessica Morley, Luciano Floridi, Libby Kinsey and Anat Elhalal · May 15, 2019 preprint; revised September 13, 2019

Open reading & exercise →

We adopt only the specific points described here. These contributors’ and organizations’ broader ethical positions are not identical to Catholic teaching. A published principle is not evidence that a product always satisfies it. No affiliate recommendation determines the conclusions in this section.

07 / Read, verify, discuss

Follow the sources.

Source passages and links checked October 26, 2026. Summaries are paraphrases; scenario feedback and workflows are original OutOfAI℠ applications. Read the full context before drawing a doctrinal or practical conclusion.

References appear in a mixed order, not a ranking of importance or a grouping by tradition. Links within each guide lead directly to its sources.

For a team discussion

Read one foundation text, choose a business scenario, hear objections and prepare a six-part review memo together. Separate disagreements about facts from disagreements about principles. Decide who must investigate before acting. Record a correction when evidence changes.

Connect this with a business workflow →

Educational edition with selected readings. No completion certificate, automated compliance approval or assessment of personal holiness. For feedback or a source correction, contact OutOfAI℠.