AI and Data Transparency

• Permissions are applied before a prompt is built, not after. The AI cannot see content you have not given it access to.
• Private context and shared context are stored separately. Private context can help the AI reason with you privately, but it never becomes part of a shared summary, decision, or recommendation unless you explicitly allow it.
• Each environment is a visibility boundary. Work in one space does not leak into another's shared view.

The AI is a neutral steward, not a narrator. What it produces is labelled so you can judge it: fact, perspective, interpretation, inference, uncertainty, and unresolved. It distinguishes what someone said from what it concluded, and it never speaks as you.

• The AI preserves context because it helps you later, not simply because it can be collected.
• Sensitive or difficult material is classified at the most protective level that applies, and your own preference always beats an automatic rule.
• Sensitive memories are not resurfaced automatically with their content. When they are relevant, the product asks first. The one exception is a genuinely time-critical obligation you have allowed us to remind you about, where you get a content-free nudge that reveals no title, names, or detail — and a more protective preference of yours always overrides it.
• Corrections are preserved as history. Updating something never silently overwrites the record of what was there before.

• Invent context that was never provided.
• Declare a winner in a disagreement, unless one position is factually wrong or unsafe — and then it says which, and why.
• Replace human judgment in consequential medical, legal, financial, clinical, or safety decisions.
• Change what is true to match your tone or preferences. Personalisation can change how something is said, never whether it is accurate.

We use third-party AI providers to generate summaries, interpretations, and suggestions. Only the content needed for a given request is sent, after permission filtering. We do not sell your content, and we do not use it to advertise to you. We cannot make absolute promises about provider-side model training, because that depends on provider terms we have not independently audited — this is stated the same way in our Privacy Policy.

AI output can be incomplete or wrong. Every summary and suggestion is meant to be checkable: you can see what it was based on, correct it, and the correction is kept. If something reads wrong, tell us through Support or the in-product Give feedback option.

Questions about how the AI handles your information? Email hello@contextislove.com.

Context is Love LLC · New Jersey, United States