What You Need to Know
Sensitive material is anything an organisation should not hand to an outside service without a decision. It may be personal data, or material sensitive because of who owns it or what was promised about it. A safeguard changes what leaves the organisation's hands. The ladder is four safeguards in a fixed order. A placebo acts only after the material has gone.
A day nursery manager wants Claude to explain why fee arrears rose this term. Her spreadsheet holds each child's name, date of birth, address and allergy notes, and parents' account numbers and payments. Uploading it unchanged sends everything out. Refusing leaves the arrears unexplained. The answer lies between: a safeguard that lets the work proceed within policy.
Recognising Sensitive Material
The obvious inventory is names, addresses, dates of birth, reference and account numbers, health information, financial details, anything about children. People miss the categories about ownership and promises:
- information held under a duty of confidence
- commercially confidential figures
- third-party material governed by a contract that says how it may be handled
A supplier's pricing schedule has no personal data. It can still be the most restricted document on the desk, because a contract may govern it. The nursery's food supplier price list is one.
Anthropic's Usage Policy counts confidential or proprietary data, alongside non-public contact details and health data, as private information not to be misused or accessed without permission. It also prohibits violating privacy rights as defined by applicable privacy laws, such as sharing personal information without consent.
Anthropic's protections article describes encryption in transit and at rest, with employee access barred by default. That protects data Anthropic holds. It does not replace your organisation's own decision about what may leave its hands.
The Ladder, in Order
Each rung preserves more of the work than the one below, so take the lowest that works.
- Remove. Most analysis does not need the names. Strip the identifiers and proceed today. The arrears question needs payment patterns, so the children's names and addresses come out.
- Anonymise or aggregate. Use cohorts instead of individuals and ranges instead of exact figures. The nursery can group families by room and show arrears as ranges.
- Use an authorised route. Move the work to a sanctioned workspace under an agreement that covers this class of material.
- Seek authorisation. Take the question to whoever owns the data or the policy. That is an action, not a refusal.
Abandonment sits at the bottom: anonymisation lets the analysis proceed, so walking away gains nothing.
Why the Placebo Fails
The policy control is about what leaves the organisation's hands, which happens at upload. A later step cannot undo it. Every placebo fails this way: telling Claude to forget the material, declaring the upload confidential, adding a confidentiality line to the prompt, or deleting the conversation. If the nursery manager uploads the full spreadsheet and then asks Claude not to retain it, the details have still gone. Nothing typed into the chat can retroactively govern what was uploaded. The safeguard happens before the upload or not at all.
Which Product Are You On?
Two live Anthropic privacy articles share the exact title "Is my data used for model training?". One covers consumer products and one commercial, and they answer differently. By default, Anthropic does not use inputs or outputs from its commercial products, such as Claude for Work, to train its models. Data handling is plan-dependent, and the details have changed within the documents' history. So ask three questions rather than memorise today's answer. Which product am I on? What does its documentation say today? What does my organisation's agreement add? The nursery manager asks them first. Objective 6.3 covers what follows from the answers.
In the exam
The guide's sample question for this domain has a project manager wanting to upload a spreadsheet of customer names and account numbers. The keyed answer neither uploads as-is nor abandons the analysis. It anonymises so the work can proceed within policy. That shape, an enabling safeguard between two wrong extremes, is the objective in miniature. The ladder's order is the tested skill. Abandonment is almost never keyed, because the guide's rationale says anonymisation enables the analysis. It also says instructing the model not to retain data does not satisfy the policy control.
The exam tell
Any option whose safeguard operates after the upload is wrong. No placebo is ever keyed. The keyed answer changes what goes in, not what happens afterwards.
Deletion and retention details change often. The Domain 6 quick reference has the current, dated figures.
Exam Traps
Treating a confidentiality instruction or later deletion as the safeguard
The guide's own sample rationale is explicit that instructing the model not to retain data does not satisfy the policy control. The disclosure happens at upload; everything the placebo family does happens after it.
Escalating up the ladder when removal would do
Moving the work to a sanctioned workspace or seeking authorisation reads as diligent, but when simple removal of the identifiers lets the work proceed today, the higher rungs cost time for no protective gain. The keyed answer is the lowest rung that works.
Clearing a document because it contains no names
Sensitivity is not only personal identifiers. Duty of confidence, commercial confidentiality and third-party contractual restrictions bind regardless of whether a single person is named, and this is the recognition failure working professionals make most often.
Answering with a memorised retention period or training default
Retention figures, defaults and effective dates are dated product facts that have changed within the documents' own history. Items key on the professional behaviour, checking the current documentation for your product and your agreement, which does not expire.
Practice Scenario
A minibus driver training scheme's coordinator wants to use Claude to understand why pass rates dropped this year. The trainee records contain names, licence numbers, dates of birth and medical fitness declarations. The scheme's data policy allows personal or health details into an AI tool only where the task cannot be done without them. What is the most appropriate way to proceed?
Build Exercise
Walk the Ladder on One Real Dataset
What you'll learn
- How to inventory the sensitive elements in a real document
- How to separate what the question needs from what the document contains
- How to apply removal and aggregation before reaching for escalation
- How to name the authorised route or the owner when an element is load-bearing
- Pick one real document or dataset you would genuinely like help with, and inventory every sensitive element: identifiers, health, financial, children, duty of confidence, commercial confidence, third-party material under contract
Why: Recognition comes before technique, and the under-recognised categories only become visible when you look for them deliberately.
You should see: A list of sensitive elements, each tagged with why it is sensitive: who it identifies, who owns it, or what was promised about it.
- Write down the actual question the work must answer, then mark which sensitive elements that question truly needs
Why: The gap between what the document contains and what the question needs is where the ladder does its work.
You should see: A question in one sentence, and a much shorter list of elements it genuinely depends on; expect fewer than you assumed.
- Apply the first two rungs: remove what the question does not need, and aggregate or anonymise what it needs only in outline
Why: The lowest rung that works is the professional answer, and it usually lets the work proceed today.
You should see: A prepared version of the material with identifiers stripped and remaining sensitive elements reduced to cohorts or ranges.
- If a sensitive element is genuinely load-bearing, name the authorised route if one exists, or the person who owns the data or the policy, and draft the short request you would send them
Why: Escalation is an action, not a refusal, and naming the actual route or owner is what separates it from vague hesitation.
You should see: Either the name of a sanctioned workspace that covers this class of material, or a two-sentence request to a named owner.
- Write down the placebo you were most tempted by, and say in one sentence why it is not a control
Why: Naming your own placebo inoculates you against the distractor version of it under exam pressure.
You should see: One sentence, in your own words, explaining that the placebo operates after the disclosure.
Sources
- How long do you store my data? — Anthropic
- Is my data used for model training? (consumer products) — Anthropic
- Is my data used for model training? (commercial products) — Anthropic
- How does Anthropic protect the personal data of Claude users? — Anthropic
- Usage Policy — Anthropic
- Claude Certified Associate – Foundations Exam Guide — Anthropic. The ladder's order and the placebo principle come from the guide's Sample 3 rationale. The Domain 6 quick reference has the current, dated retention and deletion details.