Data Processing Agreement (DPA) & General Data Protection Regulation (GDPR)

This page answers GDPR-related questions about visitor chatbot data and describes technical data-management capabilities in T3AC.

Save Chatbot History

T3AC allows administrators to control whether visitor chatbot history is stored.

The Save chatbot history option is available under the AI Chatbot configuration. For this setting, see Save chatbot history.

Backend steps

  1. Open AI Chatbot in the TYPO3 backend.

  2. Click the Chatbot tab.

  3. In Settings, enable or disable Save chatbot history.

When enabled, T3AC can store chatbot-related history such as visitor conversations and generated answers.

When disabled:

  • New chatbot conversations and answers are not stored as durable chatbot history.

  • Past-chat reopen is not retained for new conversations.

  • Thumbs-up/down feedback is hidden.

  • Live chat still works. The current session is kept only in a short-lived server cache (ns_t3ac_ephemeral_chat, about 24 hours), not as a durable TYPO3 log.

Disabling chatbot history affects newly generated chatbot and feedback data. Existing records are not automatically removed. Use the history cleanup scheduler to remove previously stored records. See History cleanup.

vendor/bin/typo3 t3af:history:cleanup
vendor/bin/typo3 t3af:history:cleanup 30

The first command uses the default retention of 90 days. Passing a number (for example 30) deletes usage history older than that many days.

Data Processing Agreement (DPA) Considerations

Administrators should consider and document:

  • Visitor chatbot data storage (messages, session metadata, visitor history)

  • Identifiers: hashed IP (MD5) and browser visitor id — treat hashed IP as pseudonymised personal data, not anonymous data

  • Browser localStorage key chat_user_id (still created for live chat when history is off)

  • Ephemeral live-session cache (~24 hours) when history is off

  • Rate limiting may hash the client IP independently of the history toggle

  • Data retention and deletion

  • Data sent to the AI provider (messages + RAG context; not IP/UA/cookies in the LLM payload)

  • BYOK vs T3Planet Credits — see T3Planet Credits

  • Provider DPA / no model training

Data Processing Agreement (DPA) Questions

Question: Does T3AC process frontend visitor information?
Answer: Yes, when a visitor uses the chatbot. T3AC can store conversation text, a hashed IP, and a browser visitor id. It does not store a frontend-user (fe_user) ID or cookies as database fields.
Question: Does T3AC store visitor conversations?
Answer: Yes — when Save chatbot history is on (the default).

Table

What it holds

tx_nst3ac_domain_model_chatbot_chat

Session metadata

tx_nst3ac_domain_model_chatbot_messages

Visitor messages and AI replies; optional feedback

tx_nst3ac_domain_model_chatbot_history

Links a visitor to a chatbot session
Question: Which identifiers are stored?
Answer:

Data

Stored?

Note

Full chat messages

Yes

When history is on

IP address

Hashed (MD5)

Not stored as plain IP. Still treat as personal data.

Browser visitor id

Yes

Created in the browser (localStorage chat_user_id) and sent to the server

Frontend user (fe_user)

No

—

Cookies as DB fields

No

Browser local storage is used for the visitor id
Question: Can visitor logging be disabled while keeping the chatbot?
Answer: Yes. Turn Save chatbot history off.

When off:

  • New chats are not written to the three history tables

  • The Chatbot card in Usage Analytics is hidden

  • Thumbs and past-chat reopen are hidden

  • Live chat uses ephemeral cache (~24 hours)

  • chat_user_id is still created for the live session

Existing rows stay until deleted or purged. See Usage Analytics.

Question: Are there other places that still store visitor data?
Answer:

Place

What is stored

Browser local storage (chat_user_id)

Random visitor id, also when history is off

Ephemeral cache ns_t3ac_ephemeral_chat

Current live session only (~24 hours) when history is off

Rate limiter cache

Hashed client IP, independent of the history toggle
Question: What is sent to the AI provider?
Answer: Visitor messages plus required RAG / context and system instructions.

Not sent in the T3AC / AI Foundation payload: visitor IP, User-Agent, session ID, cookies, or the browser visitor id.

Question: How long are records retained?
Answer: Until they are deleted — manually in Usage Analytics, or by t3af:history:cleanup when a Scheduler task is set up. There is no hard expiry on each row by itself.
Question: Logging via AI Foundation?
Answer: Separate from T3AC history. AI Usage stores a SHA-256 prompt fingerprint, tokens, and timing — not the full conversation. Privacy level can reduce or stop AI Usage rows; it does not stop T3AC history.

See AI Usage & Logs and AI Providers (privacy level).

Question: Does T3AC use conversations to train foundation models?
Answer: No. Whether a vendor trains on API data is governed by the provider contract / DPA, not by a switch inside T3AC.
Question: MCP?
Answer: If AI Foundation MCP is enabled, T3AC tools can run chatbot operations through connected clients. That is backend/editor access, not public-visitor processing. Restrict MCP as an access-control topic.

See MCP Server.