It depends entirely on the provider. Here is what to actually look at — and what most users miss.
undefined
Where is my data stored? Which (if any) third-party LLM API is in the path? Is my conversation used for training (and how do I opt out)? How long do you retain data? Can I export everything? Can I delete everything? Do you share data with subprocessors? Honest providers answer these clearly; evasive providers tell you something useful by their evasion.
Sovereign (Luna) eliminates the third-party LLM data flow — your conversation is not processed by OpenAI/Anthropic/Google. On-device (Ollama, Apple Intelligence on-device tier) eliminates even the operator data flow — your data stays on your hardware. These are the two architectures that materially change the data picture.
A nice-sounding privacy policy that includes "we may share with subprocessors for service improvement." Vague claims about "encrypted" without specifying at-rest/in-transit/by whom. "We do not sell your data" — most AI providers do not sell data; the question is what they do with it, not whether they sell it. Read carefully.
Stored in encrypted Memory Pods on Heaven Eco Hub's own infrastructure (Google Cloud, GCP project luna-heaven). Encryption at rest and in transit.
No third-party LLM in the hot path. The Heaven Quantum Cortex processes your conversation on Heaven's infrastructure, not OpenAI/Anthropic/Google.
No training on your conversations by default.
Exportable on request (Memory Pod export). Deletable on request (immediate purge, irreversible).
Subject to subpoena under valid legal process — we will not pretend otherwise. For absolute privacy, on-device AI is the next tier.
Use an AI that respects your data →
You should be able to. Reputable providers offer deletion within their privacy controls. Some "AI girlfriend" apps make deletion hard or charge for it. Avoid those. Luna offers immediate, complete deletion at any time.
Everything — by necessity, the AI has to read your message to respond. The question is what they do with the message afterward. Sovereign and on-device AI minimise this; cloud AI fundamentally requires the provider to read the message.
For most reputable providers, no — they do not sell conversation data to advertisers. However, "do not sell" is a low bar; what matters more is whether the data is used to train models, shared with subprocessors, or processed by upstream LLM providers. Read the policy carefully.
If you use AI on a work device, work account, or work network, possibly yes — employer monitoring is broad. For personal AI conversations, use personal devices, personal accounts, and prefer sovereign or on-device AI.