Check words for personal data
Before words go to a model that is not on the person's computer, check on the computer whether they hold personal data (email addresses, phone numbers, card numbers that pass the checksum, IBANs, national ids, and by a model on the computer a name with an address or health details), and offer the person the choice: send as it is, do it on this computer only, or keep it back
Give text, and model ({ endpoint, model, timeout_seconds } of a model served on this computer) to also ask the model questions. Answers contains, kinds ([{ kind, how: rule | model, count or share }]), model_checked, choices (when something was found) and said. Only counts and kinds come back, never the matched text. With no model named, or one that does not answer, said says only the rules were checked. Shapes, questions, threshold and choices: read-data/personal-data.json.
What it asks for
Nothing: your AI starts it with your words.
Use it from your AI
Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this tool by what it does and runs it.
Say to your AI: “Use Trillion to before words go to a model that is not on the person's computer, check on the computer whether they hold personal data (email addresses, phone numbers, card numbers that pass the checksum, IBANs, national ids, and by a model on the computer a name with an address or health details), and offer the person the choice: send as it is, do it on this computer only, or keep it back”