You asked the question. Maybe you typed it into Google at 11 PM after letting an AI summarize your week. Maybe a friend brought it up over dinner. Either way, it's the right question to ask: What happens to my data when I use an AI chat?
The answer depends entirely on which AI company you're talking about — and whether you're using a consumer product or a commercial API. The difference is massive, and most people don't know it exists.
Consumer AI Chat vs. Commercial API: Two Very Different Worlds
When you use a free AI chatbot — ChatGPT Free, Claude Free, Gemini — you're typically agreeing to let that company use your conversations to improve their models. That's the trade: free access in exchange for your data becoming training material.
But when a company builds a product on top of an AI model using the commercial API, the rules change completely.
Amicai is built on Anthropic's commercial API (Claude). That distinction matters more than almost anything else in this conversation.
What Anthropic Actually Says About API Data
This isn't speculation. These are direct quotes from Anthropic's published legal and policy documents:
From Anthropic's Commercial Terms of Service:
"Anthropic may not train models on Customer Content from Services."
That's not buried in fine print. It's a contractual commitment. When a company like Amicai sends your (anonymized) conversation data to Claude for analysis, Anthropic is legally prohibited from using that data to train their AI models.
From Anthropic's Privacy Center:
"By default, we will not use your inputs or outputs from our commercial products (e.g. Claude for Work, Anthropic API, Claude Gov, etc.) to train our models."
On data retention, from Anthropic's data retention policy:
Anthropic will "automatically delete inputs and outputs on our backend within 30 days of receipt or generation."
And for companies that need even stronger guarantees, Anthropic offers Zero Data Retention (ZDR), which ensures data is not stored at all after the API response is returned.
On data ownership, from the Commercial Terms:
Customer "retains all rights to its Inputs" and "owns its Outputs." Anthropic "disclaims any rights it receives to the Customer Content."
On privacy scope, from Anthropic's Privacy Policy:
"This Privacy Policy does not apply where Anthropic acts as a data processor and processes personal data on behalf of commercial customers using Anthropic's Commercial Services."
In other words: commercial API customers operate under a completely separate — and more protective — legal framework than consumer users.
But What About the September 2025 Changes?
You might have heard that Anthropic updated its consumer terms to allow training on user data by default. That's true — for free and paid consumer plans (Claude Free, Pro, and Max).
But Anthropic explicitly stated in their announcement:
"These updates do not apply to services under our Commercial Terms, including Claude for Work, Claude for Government, Claude for Education, or API use."
The consumer changes don't touch API customers. Period.
How Amicai Adds Another Layer of Protection
Even with Anthropic's strong API data commitments, Amicai doesn't stop there. Before any of your message data reaches Claude's API, we add our own privacy layer:
Your data is anonymized before it leaves our servers. When Amicai sends conversation data to Claude for analysis, it sends relationship patterns — not raw identifying information. The AI sees things like conversation topics, emotional tone, and communication frequency. It doesn't need to know who specifically you're talking to in order to generate meaningful insights.
No phone numbers reach the AI. Your contacts are referenced by first names in AI prompts. Phone numbers, which serve as identifiers in your message data, are stripped before any AI processing occurs. Claude never sees a phone number. We wrote a detailed breakdown of how phone number protection works across every layer of the system.
No raw messages are stored in our database. Your actual text messages are processed for insights, and then the analysis is stored — not the messages themselves. We don't maintain a copy of your conversations.
Your contacts never become Amicai users by proxy. The people you text don't have accounts, don't receive notifications, and aren't identifiable in our system. They exist as first-name references in your personal relationship insights — nothing more.
How This Compares to Other AI Companies
For context, here's how the major AI providers handle commercial API data:
Anthropic (Claude): Does not train on API data — contractually prohibited. Default retention: 30 days (Zero Data Retention available). Customer owns all inputs and outputs.
OpenAI: Does not train on API data (since March 2023). Default retention: 30 days. Customer owns outputs.
Google (Gemini API): Does not train on paid API tier data. Retention varies by service. Customer owns outputs.
The industry has largely converged on not training on API data. But the legal specificity and transparency of Anthropic's commitments — particularly the Commercial Terms language — is notably clear.
What This Means for You
When you use Amicai to understand your relationships better, here's the actual data journey:
- Your messages are processed locally on Amicai's secure servers
- Anonymized patterns (not raw messages) are sent to Claude's API
- Claude generates insights — your daily reflection, relationship observations, journal prompts
- Anthropic deletes the API data within 30 days — and is contractually prohibited from training on it
- The insights live in your Amicai account — the raw data doesn't persist anywhere
No AI company is reading your texts. No model is being trained on your conversations. No phone numbers are floating around in a training dataset somewhere.
The Bottom Line
"What do AI companies do with my data?" is the right question. The answer, for products built on commercial AI APIs like Amicai, is: significantly less than most people assume.
Your data isn't training material. It isn't stored indefinitely. It isn't shared with anyone. And in Amicai's case, it's anonymized before it even reaches the AI — meaning even the theoretical risk is minimized to the point of irrelevance.
The relationship insights you get are real. The privacy protections behind them are documented, contractual, and verifiable. You don't have to take our word for it — you can read Anthropic's Commercial Terms yourself.
Your conversations are yours. They should stay that way. And when you're ready to see what they reveal, here's what happened when one person let AI analyze 90 days of texts.