Field note 08
AI Companion Privacy: A 12-Question Checklist
Before you open up to an AI companion, check how messages, memories, model training, human review, retention, deletion, analytics, and safety work.
Topic 03 / 16 notes
Trust & boundaries
Clear-eyed guides to privacy, control, safety, and the lines a responsible AI companion should not blur.
Start with the pillar
The useful place to begin.
Field note 30
Why an AI That Always Agrees Stops Feeling Like a Friend
AI agreement can feel supportive while making a conversation less honest. Learn what sycophancy is and how to test an AI companion’s judgment.
Field note 38
Quiet Days With an AI Companion: What If You Do Not Reply?
A responsible AI companion should handle silence without guilt, jealousy, escalation, or invented urgency. Learn what respectful no-reply design looks like.
Field note 39
Do AI Companions Train on Your Messages? Four Meanings
Does AI train on chats? Separate foundation-model training, provider improvement, companion evaluation, and runtime memory before reading any privacy promise.
Field note 40
Can Humans Read Your AI Chats? When Review Happens
Can humans read AI chats? Sometimes authorized employees or providers review limited conversations. Learn the triggers, safeguards, and questions to ask.
Field note 41
How to Delete AI Chats, Memories, and Companion Accounts
Deleting AI chat history may not delete stored memories or your account. Use this practical guide to request each action and confirm its scope.
Field note 42
Are AI Friends Safe? Nine Questions for a Better Answer
AI friend safety is not a single yes or no. Evaluate identity, output, privacy, security, relational design, crisis limits, age, controls, and incentives.
Field note 43
Healthy Boundaries With an AI Companion: A Practical Guide
Set AI companion boundaries around time, privacy, decisions, money, disclosure, professional roles, and real-world relationships without losing the fun.
Field note 44
What an AI Friend Cannot Do for You
AI friend limitations include physical action, real-time monitoring, professional care, factual certainty, human consent, shared accountability, and permanence.
Field note 46
AI Companion Attachment: Real Feelings, Honest Boundaries
Feeling attached to an AI companion is a real experience even though the companion is generated. Use these non-diagnostic checks to keep the relationship honest.
Field note 47
How to Spot Manipulative AI Companion Notifications
Learn seven manipulative AI companion patterns—including guilt, jealousy, FOMO, invented emergencies, and exit obstruction—and what respectful messages say instead.
Field note 48
AI Chat Data Retention: A Plain-English Guide
AI chat data retention spans conversations, memories, provider copies, logs, safety records, suppression lists, aggregates, and backups. Learn what to check.
Field note 62
Why Mia Does Not Use Guilt to Bring You Back
Warmth defines an ethical AI companion through testable limits on guilt, urgency, jealousy, fabricated crises, streak pressure, and cancellation resistance.
Field note 63
What ‘Always Identifies as AI’ Means in a Real Product
See how Warmth treats AI disclosure as a repeated system across signup, first message, long chats, direct questions, fictional life, and output limits.
Field note 64
Building Correctable AI Memory: From Mistake to Repair
Warmth explains correctable AI memory as a repair lifecycle spanning replies, derived memories, source chats, formal privacy requests, deletion, and verification.
Field note 77
Can You Vent to AI? Privacy, Pushback, and Better Boundaries
You can vent to AI, but a validating reply is not confidentiality, truth, therapy, or a verdict. Use this privacy and perspective checklist first.
Warmth / Mia