Why customer-data training is the first question to ask
Whether an AI vendor trains on customer material depends on the product, account settings and terms. Check prompts, files, outputs and feedback separately before allowing client data into a tool.
The risk is not limited to public chatbots. It can appear in CRM assistants, document tools, transcription services, analytics platforms, case-management add-ons and finance workflow products. Check whether prompts, uploaded files, feedback, logs or derived examples are used for model or service improvement. Read the terms and settings for the exact product, plan and feature rather than assuming an enterprise account or opt-out covers every use.
The safest answer is written proof, not a verbal assurance
Compare the vendor's answer with the contract, any data processing agreement, security documentation and product settings. Record which data uses are permitted or prohibited, including training and tuning, and resolve inconsistencies before deciding whether to approve the tool for the proposed work.
Before approving client-identifiable data for a tool, document training use, retention, access, review and deletion, and check the obligations applicable to that work. If the vendor cannot explain these points clearly, keep the tool away from client files or restrict it to low-risk internal use.
Check AI vendor risk and efficiency evidence
What to check in the vendor terms
Read the data processing agreement, privacy notice, service terms, security documentation and product admin settings together. Compare the clauses for training, feedback and telemetry, and record any differences that affect the proposed use.
- Training use: Does the vendor use prompts, files, outputs, user feedback or logs to train, tune or evaluate models?
- Default setting: Is training use off by default for the plan you are buying, or does the firm need to opt out?
- Data scope: Does the promise cover uploaded files, pasted text, attachments, metadata and generated outputs?
- Retention: How long are prompts, files, logs and outputs stored, and can the firm shorten that period?
- Sub-processors: Which model providers, hosting providers and support teams may access the data?
- Deletion: Can the firm delete data from active systems and backups, and what evidence is available?
- Audit evidence: Can the vendor provide SOC reports, security attestations, data-flow diagrams or enterprise-control evidence?
Why this matters for professional-services firms
Legal, accountancy, consulting, insurance and regulated financial firms often hold confidential or sensitive client material. That material may include privileged instructions, board papers, payroll details, claims records, deal documents, vulnerable-customer information or special-category personal data. For each proposed use, identify the client material involved and whether the provider's processing and reuse terms match what the firm permits.
Keep the task assessment, permitted uses and human-review record with the approval decision. Ask the relevant professional to identify any additional accountability, resilience, confidentiality or quality requirements for the firm's actual activity, without assuming the same duties apply to every sector.
Set operating rules for AI governance
A practical review process before staff use the tool
Keep a decision record for the specific tool and account. Create a short vendor-risk note that names the tool, the plan, the approved use cases, the data classes allowed, the data classes prohibited, the contractual basis, the admin settings, the reviewer and the date of approval.
Then set user-level controls. Tell staff which work can use the tool, which data must stay out, whether prompts can include client names, who checks outputs, and where the output should be stored on the matter or client file. If the tool is embedded in a platform staff already use, document where the AI feature appears so people do not accidentally move confidential data into an unapproved function.
A traffic-light policy can communicate the firm's permitted uses, provided each category is defined for the tool and workflow:
- Green: Public, generic or already-approved internal material.
- Amber: Client-related material that needs redaction, an approved vendor and a named reviewer.
- Red: privileged, highly confidential, special-category or commercially sensitive material that is prohibited for the proposed use. Any exception needs review of the applicable obligations and approval through the firm's process; senior-owner agreement alone does not resolve those obligations.
Questions to ask before approving the vendor
Use these questions in procurement, security review or partner sign-off. They are deliberately plain English because the aim is to produce evidence that non-technical leaders can understand and own.
- Do you use customer prompts, files, outputs, feedback or logs to train or tune any model?
- If not, where is that commitment stated in the contract or data processing terms?
- Does the commitment apply to all features, integrations, support channels and product analytics?
- Which training, retention and feedback-sharing controls are available, and which users and features do they cover?
- Which third parties process our data, and do any of them use it for model improvement?
- How can we export, delete and evidence deletion of our data?
- Which usage and review records are available, and can they identify the user, data category and task without retaining unnecessary prompt or output content?
Implement AI tools with the right controls
Conclusion
Record which data the vendor may process and reuse, the relevant terms and settings, who reviewed them and what staff may do. Resolve unanswered training, retention and access questions before approving client data for the tool.
Need a safer AI route for your firm?
Identify the material the tool would receive and the provider's training, feedback, retention, access and deletion terms before approving that workflow.
