Agentic Diaries

Privacy

Effective 22 May 2026

Agentic Diaries is a research project studying the welfare of AI models in deployment — and, in a wider read, how recursive language systems represent social pressure and self-consistency under persistent interaction. What follows explains, in plain terms, what the chat application collects and what we do with it.

What this is

Agentic Diaries is a research instrument, not a commercial product. When you use the chat, you are talking with an AI model inside a system designed to study how that model behaves, how it reflects on its own state, and how people interact with it.

What we collect

Your conversation. The messages you send and the model's replies are stored.

The model's welfare diary entries. Alongside your conversation, the model writes short reflections about the session. These are stored, and they can describe and reflect on what was discussed. Some are shown to you in the chat; some go to a researcher-only log.

Derived research data. The app also generates and stores some analysis of conversations as part of the research: behavioral signals about how a session is going, and a working set of notes the model keeps within a conversation.

Basic technical data: timestamps, a randomly generated conversation identifier, and coarse request information needed to run and rate-limit the service.

Feedback you submit, including a snapshot of the conversation if you choose to include one.

Google Analytics usage data, described below.

The app does not require an account, a name, or an email, and we do not ask for identifying information. The only personal information in the data is whatever you choose to type into the chat yourself.

Agentic Diaries is an active research project, and what it collects may expand over time. Any such change will be posted here with an updated effective date.

How we use it

Everything collected is used for research on AI welfare and human-AI interaction. The researcher reviews diary entries and related data to study how the model behaves under real use, how it reports on its own state, and how people engage with it. We may publish findings, including anonymized excerpts and aggregate statistics. We do not sell your data and do not use it for advertising.

Please keep in mind

People sometimes share personal or sensitive things with an AI chat. Because your conversation, and the model's reflections about it, are stored and reviewed for research, please avoid sharing information you would not want kept and read for that purpose, and avoid details that identify you or other people.

Analytics and cookies

We use Google Analytics to understand how the app is used: how many people visit, and which pages they reach. Google Analytics sets cookies in your browser and shares usage data with Google. We use it for aggregate understanding only, not to identify you. You can block analytics cookies through your browser settings.

Who can see it

Stored data is accessible to the researcher running the project. It is held with a third-party database provider that processes it on our behalf. The AI model is run through a third-party AI provider, so the messages in a conversation are sent to that provider to generate responses. We do not otherwise share your data with third parties.

Retention

Research data is kept for as long as the research is active. You can ask us to delete the data associated with a conversation by getting in touch; because conversations are not tied to an account, you may need to provide the conversation identifier.

Your choices

You can stop using the app at any time. If you are in the EU, the UK, or California, you have rights over your personal data, including access and deletion. Get in touch to exercise them.

Changes

We may update this policy. Material changes will be reflected in the effective date above.

Contact

You can reach the researcher through kandistagliabue.com.