AI policy

What the assistant does

Explain suspicious messages, suggest steps and help word a neutral check question. It is not human, legal or financial advice.

What the assistant never does

Approve or reject requests, change sign-in or security settings, manage keys, pair contacts, unlock accounts, change subscriptions or perform admin tasks. A suggested question only fills in a form; you send it yourself.

Consent and switching off

The assistant is off by default. You switch it on yourself after reading the explanation, and switch it off again at any time in Settings. Core features work without the assistant.

Which data and which provider

• We only send your question and the last few messages of that conversation, after removing email addresses, phone numbers, account numbers, codes and links. The assistant does not see your requests, contacts or pairings.
• Processing via Lovable AI with a language model from OpenAI. We ask the provider not to store the content. Whether the provider uses data for training, and the region, will be verified through the contract before launch (STILL TO BE VERIFIED).

History

The conversation only exists in your browser window and disappears when you close it; ‘Clear conversation’ deletes it immediately. We only keep a count without content (90 days).

Limitations and protection

AI can make mistakes. We treat pasted text as untrusted (protection against prompt injection), limit length, remove markup and forbidden claims, and limit the number of questions. An administrator can switch the assistant off remotely.