Loopa's AI features are powered by '''Anthropic's Claude models''', reached through Anthropic's API. Before any of your information goes to them for the first time, Loopa shows you exactly what is about to be sent.
== The sheet you see before it sends ==
The first time you use an AI feature, a sheet appears with five parts:
- '''What we're sending''' — the actual list, built from your own record
- '''Why''' — what the feature needs it for
- '''Who processes it''' — Anthropic, with a link to their terms
- '''Is it kept''' — by us, and by them
- '''Is it used for training''' — no, and why that is a contractual answer rather than a promise
Two buttons: '''Continue''' and '''Not now'''. Choosing Not now cancels whatever you were doing and costs you nothing; we do not ask again in that session and there is no penalty of any kind.
'''The list is yours, not a generic one.''' It is assembled from what you have actually entered. If you have never recorded a medication, the sheet does not tell you your medications are being sent, because they are not. If you have, it says so plainly — because for a feature like the coach, they are.
You can tick '''Don't show this again for this feature'''. That holds until what we send changes. If we add a field to a feature, change the model behind it, or the terms we quote change, the sheet comes back — you agreed to a specific statement, and a changed statement is a new statement.
The small '''information icon''' beside every AI feature reopens the same sheet whenever you want it.
== What we tell you about retention, and why it is worded that way ==
Anthropic deletes what we send '''within 30 days'''. If a request is flagged for breaking their usage rules, they keep it for '''up to 2 years'''.
You will notice we do not say "Anthropic keeps nothing". Some companies have a zero-retention agreement with their AI provider. '''We do not have evidence that we have one''', so we do not claim it. If that changes we will say so, and the sheet will change with it.
'''Training is a firmer answer.''' Anthropic's commercial terms bar them from training their models on anything sent through their API. That is a contract, not an assurance, which is why it is stated flatly.
== Turning it off ==
Every AI feature has its own switch. On '''More → Preferences → Your Data''', under "What leaves your device when you use AI", open any feature and choose '''Stop using AI for this feature'''.
This is enforced on our servers, not just hidden in the app. Once you have switched a feature off, a request for it is refused — an out-of-date copy of the app cannot use it by accident. You can turn it back on the same way, and you will see the sheet again before anything is sent.
Everything in Loopa that is not an AI feature works exactly the same whether these are on or off.
== The one place AI runs without you asking ==
If you are '''new to Loopa''', the text of a forum thread or reply you post is checked against the Community Guidelines before it publishes. That is the only AI that runs without you starting it, it applies to posts and nothing else, and it stops once you are an established member. Your health log, your messages and your photographs are never screened this way. See [[Community rules and reporting]].
== What AI output is, and is not ==
It is generated by a language model. It can be wrong. '''It is not medical advice''', and it is not a substitute for a clinician — do not use it to diagnose, treat, or change a medication. The coach makes suggestions; it decides nothing about you, cannot change your subscription or your access, and you are free to ignore it.
If you would rather no health information reached an AI provider at all, do not use the AI features. Everything else in Loopa works without them.
== See also ==
- [[Your data screen|The Your Data screen]]
- [[Your data and privacy]]
- [[What your coach knows about you]]
- [[Community rules and reporting]]
Loopa's AI features are powered by '''Anthropic's Claude models''', reached through Anthropic's API. Before any of your information goes to them for the first time, Loopa shows you exactly what is about to be sent. == The sheet you see before it sends == The first time you use an AI feature, a sheet appears with five parts: * '''What we're sending''' — the actual list, built from your own record * '''Why''' — what the feature needs it for * '''Who processes it''' — Anthropic, with a link to their terms * '''Is it kept''' — by us, and by them * '''Is it used for training''' — no, and why that is a contractual answer rather than a promise Two buttons: '''Continue''' and '''Not now'''. Choosing Not now cancels whatever you were doing and costs you nothing; we do not ask again in that session and there is no penalty of any kind. '''The list is yours, not a generic one.''' It is assembled from what you have actually entered. If you have never recorded a medication, the sheet does not tell you your medications are being sent, because they are not. If you have, it says so plainly — because for a feature like the coach, they are. You can tick '''Don't show this again for this feature'''. That holds until what we send changes. If we add a field to a feature, change the model behind it, or the terms we quote change, the sheet comes back — you agreed to a specific statement, and a changed statement is a new statement. The small '''information icon''' beside every AI feature reopens the same sheet whenever you want it. == What we tell you about retention, and why it is worded that way == Anthropic deletes what we send '''within 30 days'''. If a request is flagged for breaking their usage rules, they keep it for '''up to 2 years'''. You will notice we do not say "Anthropic keeps nothing". Some companies have a zero-retention agreement with their AI provider. '''We do not have evidence that we have one''', so we do not claim it. If that changes we will say so, and the sheet will change with it. '''Training is a firmer answer.''' Anthropic's commercial terms bar them from training their models on anything sent through their API. That is a contract, not an assurance, which is why it is stated flatly. == Turning it off == Every AI feature has its own switch. On '''More → Preferences → Your Data''', under "What leaves your device when you use AI", open any feature and choose '''Stop using AI for this feature'''. This is enforced on our servers, not just hidden in the app. Once you have switched a feature off, a request for it is refused — an out-of-date copy of the app cannot use it by accident. You can turn it back on the same way, and you will see the sheet again before anything is sent. Everything in Loopa that is not an AI feature works exactly the same whether these are on or off. == The one place AI runs without you asking == If you are '''new to Loopa''', the text of a forum thread or reply you post is checked against the Community Guidelines before it publishes. That is the only AI that runs without you starting it, it applies to posts and nothing else, and it stops once you are an established member. Your health log, your messages and your photographs are never screened this way. See [[Community rules and reporting]]. == What AI output is, and is not == It is generated by a language model. It can be wrong. '''It is not medical advice''', and it is not a substitute for a clinician — do not use it to diagnose, treat, or change a medication. The coach makes suggestions; it decides nothing about you, cannot change your subscription or your access, and you are free to ignore it. If you would rather no health information reached an AI provider at all, do not use the AI features. Everything else in Loopa works without them. == See also == * [[Your data screen|The Your Data screen]] * [[Your data and privacy]] * [[What your coach knows about you]] * [[Community rules and reporting]]
