← All insights

News analysisLens: United States3 min read

Google Vault can now export Gemini conversations from Workspace. Your AI chats are becoming business records

Admins can now filter and export Gemini in Workspace conversations from Gmail, Docs, Sheets, Slides, Drive and Meet for eDiscovery. Retention rules and legal holds come later, and that matters.

Listen to this article · 5 min

AI-generated narration of the full article.

A vintage reel-to-reel tape recorder, in La Madre duotone, beside the words Chats are records
Photo: Daderot (Wikimedia Commons, CC0)

The first time an AI conversation matters in a lawsuit, nobody will ask whether the model was accurate. They will ask what the employee typed, what the assistant answered, which documents it cited, and whether the company kept all of it.

Since October 22, Google Workspace administrators can answer part of that. Google Vault, Workspace’s eDiscovery and retention tool, now supports filtered exports of Gemini in Workspace conversations, the ones that happen inside Gmail, Drive, Docs, Sheets, Slides and Meet.

What the export contains

  • Scope by person and time. Exports can target specific accounts or organizational units of up to 5,000 users, over a date range with time zone selection.
  • Full context. Each export includes the user’s prompts, Gemini’s responses, generated attachments such as images, and links to the Workspace files the conversation referenced, such as Drive documents or Gmail threads, along with citations.
  • A structured format. Exports are JSON, with metadata aligned to the format Google already uses for the standalone Gemini app.
  • Availability. The feature is in the Vault interface now for Rapid and Scheduled Release domains on Business Plus, Enterprise Standard and Plus, Education Fundamentals, Standard and Plus, Frontline Standard and Plus, and Enterprise Essentials Plus. Vault API support is due within 14 days of the interface rollout.

Two limits shape what the export can prove. It captures conversations only when conversation history is turned on. And retention rules and litigation holds for Gemini in Workspace “will follow in future releases.” Today, administrators can find and export these conversations; they cannot yet place them on hold through Vault.

AI conversations are records now

This update is small in engineering terms and large in governance terms. It confirms what records managers have suspected since assistants arrived in email and documents: a prompt and its answer are business communication. An employee who asks Gemini in Gmail to draft a reply to a customer complaint, or asks it in Sheets to explain a variance, has created a record of what they knew and how they reasoned.

Microsoft reached the same point earlier. Its Purview retention policies already cover Copilot interactions and some third-party enterprise AI apps, with copies kept for eDiscovery. With Vault catching up, both major productivity suites now treat AI conversations as content that compliance teams search. The question for most companies is whether their own records policy has caught up.

The gap between export and hold

The missing piece in Google’s update is the one legal teams care most about. When litigation is reasonably anticipated, a company has to preserve relevant information. If Gemini conversations can be exported but not yet held, preservation depends on two settings working together: conversation history must be on, and the history retention and deletion settings in the Admin console must keep conversations long enough. Google points administrators to organizational-unit settings that can stop users deleting conversations manually or delete them automatically after a period of inactivity.

That makes the Admin console, not Vault, the preservation control for now. A default that deletes conversations after inactivity is reasonable privacy hygiene and a problem the day a hold is needed.

What to do now

  1. Decide whether Gemini conversations are records in your retention schedule, and for which roles. Legal, finance and HR conversations may deserve different treatment from general use.
  2. Check the conversation history settings by organizational unit. If history is off, there is nothing to export. If automatic deletion is on, know the period.
  3. Write the interim hold procedure: until Vault holds arrive, which settings change for a custodian under hold, who changes them, and how that is documented.
  4. Test an export on a pilot group before anyone needs one, and confirm your review platform can read the JSON and follow the file references.
  5. Tell employees plainly that AI conversations at work can be retained and reviewed, the same way email is.

Google framed the update as an AI compliance capability. Read it as a signal for the whole program: the moment AI moves into everyday tools, its conversations join the company’s records, with the same obligations and the same risks as everything else employees write.

Have an AI use case stuck between prototype and production?

Tell us what you’re trying to ship. We’ll reply with honest next steps.

Discuss a use case