Skip to content
Confidence: Medium

Gemini retention rises from 72 hours to two months with history off

4 of 4 claims confirmed · 3 independent origins

What happened

Google is extending retention of Gemini conversations made with chat history disabled from 72 hours to around two months, according to TechCrunch and Ars Technica. The rollout also separates chat history from model-training permissions, allowing users to save conversations while declining their use for training—a change also reported by 9to5Google. Those controls do not eliminate data retention: Google says the longer storage period supports responses, feedback processing and safety.

Evidence plate: each square is a claim, filled when confirmed (4 of 4); each dot is an independent origin (3).

What we know

Ars Technicaarstechnica.comTechCrunchtechcrunch.com9to5Google9to5google.com
  • Google will retain temporary Gemini chats and conversations made with history disabled for around two months, up from 72 hours, according to TechCrunch.
  • Users will be able to keep Gemini chat history while separately disabling generative AI training, according to Ars Technica and TechCrunch.
  • Both new settings will default to off for users who previously disabled Keep Activity, according to TechCrunch.
  • Google's updated privacy guidance says temporary conversations are not used to train AI models, according to TechCrunch.
  • The changes are rolling out globally over the coming weeks, according to Ars Technica and TechCrunch.

What we don't know yet

  • Google did not tell TechCrunch what specifically prompted the retention increase or whether it would change how conversations are accessed or reviewed.
  • The reports do not establish when every Gemini user will receive the new controls.

Why it matters

Gemini users who disable history will have their conversations stored for around two months instead of 72 hours, even as separate controls let them decline model training without losing saved chats.

Claims

Related coverage

How this story was checked

Credibility score 84 out of 100

Source map · 3 outlets / 3 articles

How the score is built

  • Source reliability
  • Corroboration
  • Primary source
  • Claims checked against the source
  • No contradiction
  • Settled
  • Claim attribution
  • AI disclosure

Checked claims were verified against the cited source text. Weights are fixed and explainable.

Human accountability

  • Drafted by an automated system. High-risk topics require a human editor's approval before publication.