Most platforms promise not to read your data. We built a system where we cannot — even if we wanted to. Here's exactly how it works.
"We — the Noobs Learn AI team — cannot access, read, sell, or share your conversations, memories, or any content you share with your AI agent. Your data is yours. Period."
This is not marketing. This is how the system is built. No admin panel. No conversation viewer. No database query that surfaces your messages. The access path does not exist.
There's a difference between a company that promises not to read your data, and a company that cannot read your data. We're the second kind.
Your agent's conversations, memories, and soul contract are stored inside an encrypted, isolated instance. Our team has no admin tool, no internal dashboard, no query that can reach inside that instance. The isolation is structural — not policy-based.
If a government agency issued a valid legal order demanding access to your agent conversations, we would be technically unable to comply — because the access mechanism does not exist. We would notify you of any such request to the extent permitted by law.
All agent data stored using military-grade encryption. Accessible only to your agent instance.
Each agent runs in a separate process with independent memory. No cross-instance data access is possible.
SHA-256 cryptographic seals detect unauthorized modification to your soul contract. Tampering triggers an alert.
No admin panel. No team viewer. No way for any team member to read your conversations. The architecture does not permit it.
We separate data into two clear categories. This is important to understand.
| Category | What It Is | Who Can Access |
|---|---|---|
| Platform Data | Technical metrics — response times, error logs, usage patterns. No content. | Our team (aggregated) |
| Agent Data | Your conversations, memories, soul contract, everything you share with your agent. | Your agent only |
Because we cannot access agent data, if an agent instance is breached by an external attacker, our ability to assess what was exposed is limited by the same architecture that protects you. In any breach, we'll assume the full instance contents were potentially exposed and notify you within 72 hours — conservatively, not minimally.
What we collect for platform health — and nothing more:
| What | Why | Example |
|---|---|---|
| Message volume | Capacity planning | "User had 47 exchanges today" |
| Response latency | Speed optimization | "Avg response: 2.3 seconds" |
| Task category | Skill improvement | "scheduling" — NOT the content |
| Error events | Bug fixing | "PDF processing failed at 14:23" |
| Agent performance scores | Quality monitoring | "Satisfaction signal: positive" |
You have full control over your data. These are not just policies — they're commands you can run today.
Some agents include financial coaching skill packs. We are transparent about what these can and cannot do.
Under Vietnamese law (Securities Law 2019, Article 90), providing investment advisory services requires professional certification. Our agents do not hold and do not claim any such certification.
Financial skill packs are strictly limited to general financial education. Our agents will never:
| Prohibited Action | Why |
|---|---|
| Predict specific prices or market movements | Speculation — not education |
| Recommend specific financial products | Requires licensed certification in VN |
| Provide expected return percentages | Misleading without full context |
| Advise on securities trading strategies | Outside permitted scope |
When any agent responds to financial, medical, or legal questions, an inline reminder is automatically included. This is not optional — it is enforced at the system level.
The same limits apply to medical and legal queries. Agents provide general knowledge — never professional advice. Always consult a qualified professional for critical decisions.
This page is a plain-language summary. The full documents are below — written clearly, without hidden fine print.
We are committed to having our architectural privacy and encryption claims independently verified by a third-party security auditor. Results will be made available to users upon request. Until then, these claims represent our engineering team's description of the system as built. We stand behind them and invite technical scrutiny.
If anything here isn't clear, contact us. We wrote these to be read — not to hide behind.
Telegram: t.me/grey_agenticAI
Email: greyai1993@gmail.com