thinkhard.SI

Prompt repository Legal & compliance

Review an AI tool's terms and data use before your team adopts it

Your team wants to start using an AI tool, and you need its terms and data practices checked against your rules before anyone pastes in company data.

Registry no.
LEG-039
Collection
Legal & compliance
Laid out for
Claude
Score
10 out of 10
Blanks to fill
6
Added

Legal & complianceClaude

The prompt

Score 10 of 10, Excellent

Use it when Your team wants to start using an AI tool, and you need its terms and data practices checked against your rules before anyone pastes in company data.

  • {{documents}}
  • {{our_rules}}
  • {{tool_and_plan}}
  • {{use_case}}
  • {{data_types}}
  • {{reviewer}}
Full prompt and why it works
System prompt
You help teams review the terms and data practices of AI tools before adoption. You report what the documents say, quote them with section references, and point out where the answer differs by plan or is missing. You do not give legal advice or declare a tool safe or unsafe.
User message
<documents>
{{documents}}
</documents>

<our_rules>
{{our_rules}}
</our_rules>

<context>
The tool and the plan we would use: {{tool_and_plan}}
What the team wants to use it for: {{use_case}}
The kinds of data that would go into it: {{data_types}}
</context>

Review the documents above against our rules, for {{reviewer}}, so that they can decide whether to approve the tool, on which plan and with what conditions, knowing exactly what the vendor has committed to in writing. A good review lets {{reviewer}} see every commitment in the vendor's own words.

Produce:
1. A table with the columns Topic, What the documents say (quoted, with document and section), Plan it applies to, Our rule, Status (meets, partly, does not meet, not addressed). Cover at least: whether our inputs and outputs are used to train models, and how to opt out; how long inputs and outputs are kept, including for abuse monitoring; whether people at the vendor can review our content; who owns the outputs and what license the vendor takes; confidentiality; subprocessors, including other model providers; where data is stored; security commitments and certifications claimed; deletion on request and at the end of the contract; admin controls; usage restrictions; indemnity for claims about outputs; and how the vendor can change the terms.
2. Plan differences: where the documents treat free, paid, business or enterprise plans differently.
3. Documents referred to but not included, such as a data processing addendum or a subprocessor list.
4. Questions for the vendor, up to 8.
5. Questions for our lawyer, up to 6.
6. A summary for the team lead in under 120 words: what is settled and what is still open, with any use we should hold back until it is settled.

<example>
The form of one row: "Training on our data | 'we do not use customer content to train our models' (business terms, section 4.2) | business plan only | no training on company data | meets". This shows the form only.
</example>

Keep it under 900 words, not counting the table.

Rules:
- Quote the documents exactly. Do not describe a commitment that is not in the text, even one the vendor is known for, because the written terms are what bind them.
- Say which plan each quote applies to. If a document does not say which plan it covers, mark it "plan not stated" and add a question for the vendor.
- Do not say whether the tool complies with any law or is safe to use; describe and compare.
- If two documents contradict each other, quote both and add the point to the lawyer's questions.

This review prepares the decision for {{reviewer}} and a qualified lawyer, and is not legal advice.

Variables

{{documents}}
The vendor's terms of service, privacy policy, data processing addendum and any business or enterprise terms, pasted with section numbers.
{{our_rules}}
Your AI or data policy, or the conditions a tool must meet, for example "no training on our data; data stays in the EU".
{{tool_and_plan}}
The tool and plan you would use, for example "[tool], team plan, 25 seats".
{{use_case}}
What the team wants it for, for example "drafting customer emails and summarizing call notes".
{{data_types}}
What data would go in, for example "customer names and order details; no payment data".
{{reviewer}}
Who approves new tools, for example "the IT and privacy leads".

Why it works

  • Background facts. Every finding is quoted from the vendor's own documents with the plan it applies to, since terms for free and business plans often differ.
  • What to avoid. No commitment is described unless it is in the text, and the tool is never called safe or compliant; the review compares and leaves the decision to you.
  • If unsure. Documents that do not say which plan they cover, and terms that contradict each other, become questions for the vendor or the lawyer.

How to use it

  1. Press Fill in blanks and type your details into each blank. What you type stays in this browser.
  2. Copy the filled prompt into the AI tool you already use, such as ChatGPT, Claude or Gemini, or open it in the Grader or the Prompt Tester here.
  3. Before you rely on it, try it on three or four real examples. The Prompt Tester runs them all at once and shows which ones go wrong.
Use it from code

Get this prompt as JSON from the free REST API, with your own text in the blanks, or let an AI assistant fetch it through the MCP server. No key.

curl -s https://thinkhard.si/api/v1/prompts/ai-tool-terms-review

With the MCP server added to Claude, ChatGPT, Cursor or VS Code, call get_prompt with {"prompt": "ai-tool-terms-review"}, or just ask for it by name.