Gemini unbundles chat history and training, extends retention | Magica
Gemini separates chat history from AI training, but keeps history-off chats longer
Editorial Team
••📖6 min read
Google will let Gemini users save conversations while opting out of generative AI training on future chats. The global rollout also extends storage of history-off and temporary chats from 72 hours to around two months; human review and longer retention can still apply.
Gemini users will be able to save conversations while opting out of generative AI training on future chats.
Google will retain history-off and temporary chats for around two months, up from 72 hours.
Opting out of training does not stop all human review, other service-improvement uses or longer retention.
Google will let users of Gemini, its personal AI assistant, keep their chat history without permitting generative AI training on future conversations. Alongside that new choice, it is extending storage of chats that users leave out of their history—from 72 hours to around two months.
Google announced the change on November 16 and said in its announcement that the controls would roll out globally over the coming weeks. The update separates two decisions previously bundled together: whether users can return to saved conversations and whether Google can use those conversations to train models. Neither decision alone determines when Google deletes the data.
Google’s November 16 announcement pairs separate history and training choices with longer storage of history-off chats. Source: Google.
Keeping history without contributing future chats to training
Gemini's previous Keep Activity switch governed both saved history and permission to train on chat content. Users who disabled training through that setting also lost the ability to return to their saved conversations, as reporting on the new controls explains.
The new History setting governs saved conversations; a separate setting, labelled “Chat content including text & uploads,” controls generative AI training. Google's updated Privacy Hub says disabling chat-content training covers future chats, unless users submit feedback. It does not promise to undo prior training or erase previously reviewed conversations.
That content can extend well beyond typed prompts: Google lists audio transcripts, attachments, Gemini's responses, tasks and tool calls, location information, and data from connected apps among the information covered by the training setting.
Microsoft is pitching Copilot Workspaces as a shared environment for people and AI agents. Team collaboration is already in early access, but personal Autopilot integration is still coming soon, with final agent pricing unannounced.
Editorial Team
Keeping History on also preserves access to features. Google says some connected apps, including Google Workspace—which lets Gemini retrieve information from users' emails and other content—are unavailable with History off. Tasks that use a remote browser or computer also require History. Separating training permission from history therefore removes a condition attached to more than the convenience of rereading a chat.
Google's history-management instructions say History is on by default for users aged 18 and over. Saved history is automatically deleted after 18 months by default; users can choose three or 36 months, turn off automatic deletion, or delete history manually.
These instructions cover personal accounts. For work and school accounts, Google Workspace administrators control history settings, and different data-handling terms may apply.
Leaving chats out of History means longer storage
Google's revised policy says temporary chats and conversations made with History off are retained with the account for around two months. History-off conversations will not appear in the user's Gemini History.
The previous retention period was 72 hours, according to a comparison with a November 12 archived policy. The new period is approximate: Google says around two months, rather than specifying a fixed 60-day deadline.
Google frames the increase as a safety and security measure as AI systems become more capable. Its detailed notice says these chats are used to respond to users, process feedback and protect Google, users and the public, including with human reviewers. It says only the last 24 hours of a chat are used as context to respond, and cites potential system failures as a reason for retaining chats with the account.
Asked what prompted the change and whether conversation access or review would change, Google did not answer those questions and instead directed the publication to its announcement, the report says. The stated safety rationale therefore leaves the specific operational need for the longer window unexplained.
The same report offers a narrower comparison with rival assistants: OpenAI says temporary chats in ChatGPT may be kept for up to 30 days, while Anthropic retains incognito chats in Claude for 30 days by default. Those figures concern temporary or incognito conversations, rather than all data held by either company. Legal, safety and policy-investigation exceptions can extend storage.
Google’s company-reported policy allows reviewed chats and related data to remain for up to three years, even after deletion or a later training opt-out. Source: Google Privacy Hub.
The training opt-out has exceptions
Google says temporary chats are not used to train its AI models. It also says future chats made with History off are excluded from generative AI training when users do not submit feedback.
Feedback changes that treatment. With History off, submitting feedback causes Google to collect the last 24 hours of chats and their contents, including attachments and connected-app data, to help maintain and improve Gemini. Reviewed feedback, associated conversations and related data can be retained for up to three years, disconnected from the account.
Disabling the separate chat-content training setting also leaves other uses intact. Google says future chats will still be reviewed and used to improve its services in other ways and to protect Google, users and the public. Gemini settings do not control processing chats into anonymized data for service improvement.
Google says a subset of chats undergoes human review. Chats sent to service providers are separated from accounts, and reviewed chats can remain for up to three years even after users delete them or turn training off. Google also reserves longer retention for legitimate business or legal purposes, including security and abuse prevention. The roughly two-month period is therefore not a universal deletion guarantee.
Audio and Gemini Live video and screenshare recordings have a separate service-improvement control. If enabled, it covers recordings already stored in History as well as future recordings. Turning it off does not prevent training on transcripts and attachments if chat-content training remains on.
Deletion has boundaries too. Removing Gemini history does not delete data already saved by other Google services, or remote browser and code-execution data, which have separate controls.
The regulators said the revised settings addressed their concern that fear of losing history could discourage training opt-outs. That was a finding about one aspect of OpenAI's design, not approval of all its privacy practices or evidence that the investigation prompted Google's change.
For existing Gemini accounts, the immediate issue is how training permission carries over. Google says both new settings will start off for users who previously disabled Keep Activity, as coverage of the rollout also notes. According to the earlier retention-policy report, Google says users who previously had Keep Activity on will retain history and can decide separately about training.
The published accounts differ on training defaults. Google's history-management page says the chat-content and recording improvement settings are off by default; Ars Technica describes the new training control as opt-out. The retained explanations do not establish how those descriptions map to every existing account during migration.
As the controls arrive, users will need to check History, chat-content training and recording improvement separately. For Google, clearer migration instructions and an explanation of why safety requires around two months of account-linked storage would resolve the main questions its announcement leaves open.
Google is rolling out file references and sharing for Gemini skills as adult personal accounts lose the ability to use Gems. Automatic conversion preserves supported material, but file limits and missing tools can interrupt existing workflows.
Muddy Waters disclosed a short position in Nvidia while questioning a possible connection involving senior adviser Ajay K. Puri. Its own account leaves the key identity question unanswered, separating its new claims from an existing server-diversion prosecution.
Nvidia reported $115.8 billion in quarterly revenue in its first earnings release since co-founder Jensen Huang resigned. Strong demand gives interim CEO Colette Kress a financial cushion, while a campaign for Huang’s return collides with unresolved investigations and the board’s reported opposition.
OpenAI launched GPT-6.2 Astra and Sol for professional work, cutting Astra’s standard input and output token prices by 20%. The models offer cheaper paths through selected tasks, but research evaluations, reporting failures and unfinished independent safeguard testing complicate the case for broader autonomy.
Microsoft’s first live speech-to-speech model offers OpenAI API compatibility and planned enterprise integrations. Its published evaluations show a low-cost option, with several rivals completing more agentic tasks.
Canberra-based Trellis Data is offering its 220-billion-parameter Agentaus 1 model to business and government. Its pitch goes beyond keeping data in Australia, but performance, deployment costs and a planned capital raising remain key tests.
Perplexity is adding direct collaboration between frontier models to Computer, letting a lead agent coordinate up to 16 supporting agents. Company-reported benchmarks suggest mixed teams can outperform repeated copies of one model, but leave customer costs and production reliability unresolved.
Marsh reported 105 client cyber incidents in the last two weeks of October, against 48 in the final two weeks of May. The broker links the increase partly to AI, while warning that insurance panel access may not guarantee an available response team.
Meta’s Muse personal AI agent has launched in Australia, with some local shopping and booking connections to follow. Its browser and payment tools broaden its reach, while the company’s technical disclosures qualify its privacy promises.
Microsoft and Salesforce are extending their sales software alliance with Copilotforce, combining CRM workflows with Microsoft 365 context under existing permissions. The Copilot plugin enters public preview, while Teams remains in pilot; earlier Microsoft tools and a Claude beta raise the bar for proving its value.
Microsoft expands MAI Image 2.7 from an editing preview into image generation and API access, with cheaper Flash pricing and a Canvas workspace. Benchmark claims depend on the workload, while the MAI Labs interface carries restrictions on commercial use.
OpenAI is widening applications for its legal AI configuration after improving research scores and expanding its search index. Access still requires approval, and benchmark results show substantial limits alongside the gains.
Microsoft has made Word, Excel and PowerPoint creation generally available inside Copilot, combining editable files with workplace permissions and collaboration. The broader shift toward autonomous work remains staggered across previews, while its claimed cost advantage rests on company-funded testing.
Knit Health’s NEA-led Series A funds deployment of Med-BEAM for care navigation and hospital operations. Its early company benchmark beats three Claude models at predicting clinical actions, while better patient outcomes remain unproven by that test.
Science Corporation’s Series D follows its European retinal-implant launch. Clinical results support improved central vision, but U.S. approval, eligible patients and commercial terms remain consequential questions for the startup, reportedly valued at $3.5 billion.
Google is removing the subscription requirement for Gemini’s task mode in six countries, while adding U.S. payment, calling and messaging features. The expansion broadens access to delegated work, but usage limits, permissions and website compatibility still govern what it can finish.
Eligible Pixel owners can join a Play Store beta to generate home-screen widgets from descriptions, but device, language and regional restrictions still apply.
Microsoft’s new in-house coding models compete with outside suppliers inside GitHub Copilot. Company tests show lower task costs at promotional prices, while its continuing financial ties to OpenAI complicate the rivalry.
Valkryn disclosed a $15 million seed round and early military testing contracts. Its evaluator-focused Battle Lab faces existing commercial tools, public testing capabilities and funding constraints that software alone cannot resolve.