Check before company data goes in
Compare 63 AI products, identify the data in scope, and leave with an evidence-backed review receipt, not a vague safety score.
Takes about 2 minutes · 63 products · Public disclosures only
You'll get a summary like this
3 of 5 selected tools publicly state some training use.
Sample output from a five-tool selection. Yours updates as you pick products.
1 · Pick your products
63 coveredNothing you select or classify is stored.
You'll get a summary like this
3 of 5 selected tools publicly state some training use.
Sample output from a five-tool selection. Yours updates as you pick products.
From disclosure to approval
Keep six pieces of evidence
A no-training statement answers one question. Use this checklist before non-public data, connected systems, or automated actions enter the workflow.
Name the exact product and edition
Consumer, business, enterprise, free, and paid API paths can have different terms. Record the actual account and workspace used.
Map every data path
Include prompts, uploads, connected sources, retrieved passages, outputs, feedback, logs, support access, plugins, and downstream actions.
Verify training and improvement settings
Check the default, organisation controls, user controls, feedback exceptions, and whether settings apply prospectively only.
Confirm retention and deletion
Ask what is stored, where, for how long, who can delete it, and whether backups, abuse monitoring, caches, or stateful features differ.
Confirm location and subprocessors
Record processing and storage regions, cross-region routing, model providers, support locations, subprocessors, and contract change notices.
Set access, logging, and human gates
Use least privilege, approved connectors, auditable events, incident ownership, tested redaction, and human approval before consequential actions.
This is a decision aid, not a legal opinion or product certification. Public terms change; verify the linked source and your contract before approval.