What you will be able to do
- Explain Diligence as owning the outcome of AI-assisted work, and name its three sub-components
- Recognise when AI's role in a piece of work must be disclosed, and to whom
- Spot over-reliance, where a quick skim of an AI output stands in for real judgment
- Keep humans making high-stakes decisions about people when AI outputs may carry bias
- Use a sound process when organisational policy gives no clear answer
Key concept
Diligence — Diligence is the AI Fluency competency of taking full responsibility for AI-assisted work. You are answerable for whether it is accurate, whether you were honest about how it was made, and what it does to the people it affects, however much of it the AI produced.
1.Diligence: you own the outcome
The AI Fluency 4D Framework has four competencies: Delegation, Description, Discernment and Diligence. They form two loops. Description and Discernment are the inner loop: how you ask the AI for something and how you judge what it gives back. Delegation and Diligence are the outer loop, which wraps around every interaction. Delegation happens before you start, when you decide what to hand to the AI. Diligence happens afterwards, when you answer for the result. The K-12 educators course says it in one line: "Afterward you take responsibility for what comes back."
Most of the ethics in AI use sits in Diligence. The educators' version of the framework says Diligence ensures responsible, transparent, and accountable AI use. The builders' course is blunter: you own the outcome, not the output, and "AI wrote it" explains nothing and excuses nothing. Using AI does not move accountability onto the tool. If an AI-drafted report misleads a reader, the person who sent it is responsible. The model is not.
The nonprofit version of the framework splits Diligence into three sub-components. They give you a useful checklist for ethics questions, because most ethical failures with AI are a failure of one of them.
| Sub-component | What it asks of you | Typical ethical failure |
|---|---|---|
| Creation Diligence | Being intentional about which tools you use | Using an AI tool for work it isn't suited or approved for |
| Transparency Diligence | Being honest about AI's role | Presenting AI-drafted work as entirely your own judgment |
| Deployment Diligence | Verifying and vouching for outputs | Sending on an output you haven't checked |
Checkpoint 1 of 8· Match them up
Match each Diligence sub-component to what it covers
Tap a term, then the definition that fits it.
The framework divides Diligence into these three parts: choosing tools deliberately, being honest about AI's involvement, and checking and standing behind what you release.
“Transparency Diligence (being honest about AI's role), and Deployment Diligence (verifying and vouching for outputs)”Source: academy.claude.com
2.Transparency and attribution: saying how AI was used
Transparency Diligence matters whenever other people treat your work as evidence of your own expertise. In the builders' course, the pre-ship checklist asks two questions under "Responsibility": could this output be misread or misused, and have you been transparent about AI's role? The K-12 course asks educators the same two things before they finish a piece of work: would you put your name on it, and do you need to tell anyone how it was produced?
Anthropic's description of its own teaching approach makes disclosure concrete. Learners are taught how to ethically disclose AI use to colleagues, customers and other stakeholders. That means saying explicitly how AI was used to produce documents, analyses or media before sharing them. Disclosure is not one fixed sentence attached to everything. The creative-work course frames a personal AI policy around three questions: where AI fits and where it doesn't, what you disclose and to whom, and when you will revisit the policy. The disclosure owed can vary by audience. Deciding it is part of the work.
Checkpoint 2 of 8· Check yourself
A team lead has Claude draft a client-facing analysis, checks every figure carefully, and sends it without mentioning AI. The client treats it as the lead's own work. Which Diligence obligation is still unmet?
Verification covers Deployment Diligence, but being honest about how AI was used is a separate obligation owed to the people who receive the work.
“This includes explicitly stating how AI was used in the production of documents, analyses, or media before you share it with others.”Source: claude.com
Checkpoint 3 of 8· Exam question
A fintech startup wants to use Claude to help loan officers draft explanations for credit decisions before those explanations are sent to applicants. Under Anthropic's Usage Policy, what must the company do to use Claude responsibly in this high-risk context?
Correct answer: B — Have a loan officer review each explanation before sending it, and disclose to applicants that AI helped draft it
- A. Incorrect. A disclaimer alone does not satisfy the human-in-the-loop requirement; a qualified professional must still review the specific decision before it goes out.
- B. Correct. Financial services is a high-risk use case under the Usage Policy, requiring a qualified professional to review AI-assisted content before dissemination and requiring disclosure to the end user.
- C. Incorrect. Avoiding disclosure does not meet the policy's transparency requirement, and internal-only use does not remove the review obligation once content reaches applicants.
- D. Incorrect. Retroactive periodic audits are not a substitute for prior review of each decision by a qualified professional before it is disseminated.
3.Over-reliance: when AI replaces the thinking
Over-reliance is the ethical risk that is easiest to miss, because it looks like efficiency. The K-12 course names it as a dependency tension: when could relying on AI replace thinking that people need to do for themselves? It applies to professionals as well as students. When a skimmed draft goes out as a considered judgment, the AI has effectively made the decision, and the person whose name is on it has not.
The AI Fluency framework draws the line as augmentation versus automation. Augmentation means working with AI to improve your work. Automation means having the AI do the work for you. One of the mindsets Anthropic's education team teaches, alongside the reminder that today's AI is the worst you'll ever use, is to verify in proportion to the stakes. A brainstorm needs little checking. A recommendation that colleagues will spend money on needs a lot.
Over-reliance also builds up over time. The creative-work course names convergence and skill atrophy as the long-term pitfalls. The immediate pitfall is reaching for AI when you don't yet know what you want. Anthropic's materials encourage people to keep practising skills that matter to them so those skills don't atrophy, and to think about which tasks should stay with them. In Anthropic's examples, you might draft the sensitive parts of a memo yourself and leave the summary slides to AI, or use Claude for exploratory analysis but do the final checks yourself.
Checkpoint 4 of 8· Check yourself
According to Anthropic's AI Fluency materials, which of these is a long-term pitfall of AI use, as opposed to an immediate one?
The creative-work course names convergence and skill atrophy as the long-term pitfalls. Reaching for AI when you don't know what you want is the immediate one.
“Convergence and skill atrophy are the long-term pitfalls; reaching for AI when you do not know what you want is the immediate one.”Source: academy.claude.com
Checkpoint 5 of 8· Exam question
A consumer-facing startup builds a chatbot on Claude for its e-commerce site and wants to follow Anthropic's guidance on transparency. What should the company implement at the start of a customer's chat session?
Correct answer: C — A clear notice at the start of the session disclosing that the customer is talking to an AI system
- A. Incorrect. Instructing the model to impersonate a human and deny being AI directly violates the Usage Policy's prohibition on deceptive impersonation.
- B. Incorrect. A buried terms-of-service mention does not meet the requirement for clear, session-level disclosure that a customer would actually see.
- C. Correct. Anthropic's guidance for consumer-facing chatbot applications requires disclosing AI involvement at the start of the session so users know they are talking to an AI.
- D. Incorrect. Consumer chatbots are explicitly called out as needing AI disclosure at the start of the session, not exempted from it.
4.Bias in outputs used for decisions about people
Over-reliance is most harmful when the output decides something about a person: who is shortlisted, who is promoted, who gets help. The K-12 course lists bias as a core ethical tension and asks when AI might reinforce narrow perspectives, stereotypes or existing inequities. A ranking or summary of candidates can carry exactly that kind of distortion, and it looks objective while doing so.
The course's response is to commit in advance to what you will not hand over. Its template commitment reads "I will not delegate decisions related to" and gives examples: final judgment, relational feedback and high-stakes evaluation. In practice, AI can help organise and summarise information about people, but a human makes the evaluative decision. That human weighs the AI's output as one input and checks it for the distortions the course warns about. They don't simply approve it. The builders' course adds a question about fairness: who does your work not serve well? Access is a design decision, and you should check who your assumptions exclude before you treat something as finished.
Below this professional judgment there is also a hard floor. Anthropic's Usage Policy guidance for agents gives, as an example of prohibited use, collecting, compiling or analysing personal information to build profiles based on people's protected attributes, sensitive characteristics or personal circumstances. Ethics asks more of you than that policy does. But a workflow that builds such profiles is not a grey area.
Checkpoint 6 of 8· Check yourself
A manager asks Claude to rank 50 internal applicants for a promotion. Which use of the ranking best reflects the ethical guidance in these materials?
High-stakes evaluation and final judgment stay with people, and AI outputs can reinforce stereotypes or existing inequities. So the ranking informs the decision but does not make it. Disclosure alone doesn't deal with bias.
“To protect against that concern, I will not delegate decisions related to: (e.g., final judgment, relational feedback, high-stakes evaluation)”Source: academy.claude.com
Checkpoint 7 of 8· Exam question
An HR team wants to use Claude to help draft candidate rejection emails after final hiring decisions have already been made by human recruiters. Why does this still count as a high-risk use case under Anthropic's Usage Policy?
Correct answer: D — Because employment is a designated high-risk category, so review and disclosure still apply here
- A. Incorrect. Rejection emails are not legally binding contracts, and notarization is not part of the Usage Policy's requirements.
- B. Incorrect. Claude can be used for employment-related text; the policy requires safeguards like review and disclosure, not an outright ban.
- C. Incorrect. Storage duration is not the factor that determines high-risk status; the use case category itself is what triggers the requirement.
- D. Correct. Employment is one of the categories the Usage Policy designates as high-risk due to public welfare implications, so human review and disclosure requirements apply even when a human made the underlying decision.
5.When policy runs out: judgment and commitments
Rules cover only part of what you'll face. The K-12 course says it plainly: using AI responsibly is not just about following rules. It is about ethical judgment in the grey areas where policy gives no clear answer. Other subdomains cover your organisation's AI policy. This section is about what you do when that policy is silent.
The course offers a sensible default for those moments. When you're unsure whether a use of AI is ethical, pause, consult a colleague, review the relevant guidance and pilot cautiously. Its course on values suggests trying new tools only in low-risk contexts first. None of these steps needs special authority, and all of them come before anything reaches the people affected.
Commitments made in advance make that judgment easier to hold to. The same course has learners write down what they will and won't do with AI, because stated values help you keep your non-negotiables under deadline pressure. Its list of tensions to think about goes beyond accuracy: dependency, de-socialisation, bias, agency, equity of access and environmental impact. These commitments also need upkeep. Ethical AI use is ongoing, so you revisit your commitments as the people you serve, your context and your tools change.
Checkpoint 8 of 8· Check yourself
A colleague wants to use Claude in a new way that your organisation's AI policy doesn't mention. It would affect how customers are treated. What does the ethics guidance recommend first?
Silence in a policy is a grey area that calls for judgment, not automatic permission. The recommended response is to pause, consult, check guidance and pilot cautiously.
“When I'm unsure whether an AI use is ethical, I will: (e.g., pause, consult a colleague, review district guidance, pilot cautiously)”Source: academy.claude.com
Exam traps
Each one states something that sounds right. Open it to see what is actually true.
1.If the AI produced the error, responsibility for it is shared with the tool.Why is that wrong?
Diligence puts full ownership on the person who uses and releases the output, however much of it the AI wrote.
Covered in Diligence: you own the outcome
2.If you've reviewed the AI's draft, you don't need to disclose that AI wrote it.Why is that wrong?
Verification and transparency are separate parts of Diligence. Being honest about AI's role is owed to the people relying on the work even when the content is correct.
Covered in Transparency and attribution: saying how AI was used
3.An AI ranking of people is fair because it applies the same criteria to everyone.Why is that wrong?
AI outputs can reinforce stereotypes and existing inequities, so final judgment and high-stakes evaluation stay with a human.
Sources
Every claim above is drawn from one of these pages, quoted as it was written on the date shown.
- 1.https://academy.claude.com/courses/ai-fluency-for-k-12-educators/ethics-responsible-useOfficial docs
“Afterward you take responsibility for what comes back.”
↩︎ Diligence: you own the outcome“Dependency: When might reliance on AI replace thinking students need to do themselves?”
↩︎ Over-reliance: when AI replaces the thinking“Bias: When might AI reinforce narrow perspectives, stereotypes, or existing inequities?”
↩︎ Bias in outputs used for decisions about people“Using AI responsibly isn't just about following rules; it's about ethical judgment in the gray areas where policy gives no clear answer.”
↩︎ When policy runs out: judgment and commitments“Ethical AI use is continuous. Revisit your commitments as students, context, and tools change.”
↩︎ When policy runs out: judgment and commitments“Bias: When might AI reinforce narrow perspectives, stereotypes, or existing inequities?”
↩︎ Exam trap 3“To protect against that concern, I will not delegate decisions related to: (e.g., final judgment, relational feedback, high-stakes evaluation)”
↩︎ Checkpoint“When I'm unsure whether an AI use is ethical, I will: (e.g., pause, consult a colleague, review district guidance, pilot cautiously)”
↩︎ Checkpoint - 2.https://academy.claude.com/courses/ai-fluency-for-educators/ai-fluency-framework-reviewOfficial docs
“Diligence ensures responsible, transparent, and accountable AI use.”
↩︎ Diligence: you own the outcome“Augmentation (working with AI to enhance our work) is more powerful than automation (AI simply doing work for you)”
↩︎ Over-reliance: when AI replaces the thinking - 3.
“You own the outcome, not the output. “AI wrote it” explains nothing and excuses nothing.”
↩︎ Diligence: you own the outcome“Responsibility — Could this output be misread or misused? Have you been transparent about AI’s role?”
↩︎ Transparency and attribution: saying how AI was used“Access is a design decision. Check who your assumptions exclude before you call something shipped.”
↩︎ Bias in outputs used for decisions about people“You own the outcome, not the output. “AI wrote it” explains nothing and excuses nothing.”
↩︎ Exam trap 1“Responsibility — Could this output be misread or misused? Have you been transparent about AI’s role?”
↩︎ Exam trap 2 - 4.
“Creation Diligence (being intentional about which tools you use)”
↩︎ Diligence: you own the outcome“Transparency Diligence (being honest about AI's role), and Deployment Diligence (verifying and vouching for outputs)”
↩︎ Checkpoint - 5.
“Would you put your name on the final output? Do you need to be transparent with anyone about how it was produced?”
↩︎ Transparency and attribution: saying how AI was used - 6.
“We also educate learners on how to ethically disclose AI use to colleagues, customers, and other stakeholders.”
↩︎ Transparency and attribution: saying how AI was used“encourage mindfulness about continuing to practice skills that matter to you in order to prevent skill atrophy”
↩︎ Over-reliance: when AI replaces the thinking“This includes explicitly stating how AI was used in the production of documents, analyses, or media before you share it with others.”
↩︎ Checkpoint“verify in proportion to the stakes”
↩︎ Prediction - 7.https://academy.claude.com/courses/ai-fluency-for-creative-work/putting-it-all-togetherOfficial docs
“Three questions anchor the policy: where AI fits and does not, what you disclose and to whom, and when you will revisit.”
↩︎ Transparency and attribution: saying how AI was used - 8.https://academy.claude.com/courses/ai-fluency-for-creative-work/delegation-and-diligenceOfficial docs
“Convergence and skill atrophy are the long-term pitfalls; reaching for AI when you do not know what you want is the immediate one.”
↩︎ Over-reliance: when AI replaces the thinking - 9.https://support.claude.com/en/articles/12005017-using-agents-according-to-our-usage-policyOfficial docs
“Collect, compile, or analyze personal information to create profiles based on individuals’ protected attributes, sensitive characteristics, or personal circumstances”
↩︎ Bias in outputs used for decisions about people - 10.https://academy.claude.com/courses/ai-fluency-for-k-12-educators/ai-s-constitution-pedagogy-and-youOfficial docs
“Be clear about what you will and won't do with AI. Stated values help you hold your non-negotiables.”
↩︎ When policy runs out: judgment and commitments“Testing new tools only in low-risk contexts first”
↩︎ When policy runs out: judgment and commitments
Also cited
“Diligence means owning the final result fully, regardless of AI’s role.”
↩︎ Key concept