Technology
Ars Technica

Google makes it easier to use Gemini without training AI on your data

Source Entity

Ryan Whitwam

November 19, 2026
Google makes it easier to use Gemini without training AI on your data

Google is updating Gemini with separate toggles for chat history and AI training, alongside a rebranding of Spark AI to 'Tasks'. While these updates offer more user control, Google is simultaneously extending temporary chat retention from 72 hours to two months.

Evolution of Google’s Gemini Ecosystem

Google is undergoing a significant transformation in how it manages user data and AI capabilities within its Gemini ecosystem. By decoupling chat history from AI training data through new, granular toggle controls, Google is addressing long-standing user concerns regarding data privacy. Previously, users were forced into a binary choice: either sacrifice the ability to revisit past conversations or consent to having their data utilized for training future iterations of the model. This shift marks a strategic pivot toward offering users more agency over their personal information.

The Rebranding and Expansion of 'Tasks'

Simultaneously, Google is rebranding its Spark AI agent to 'Tasks'. Now powered by the Gemini 4 Hydrogen model and offered for free, this tool is designed to act as a personal digital assistant. Its capabilities range from routine administrative tasks like calendar management and meal planning to more complex actions such as auditing emails for recurring expenses. For users in the United States, this suite of features is being further bolstered by integration with Google Pay and Password Manager, allowing the AI to handle tasks such as price comparisons and inventory checks via phone calls to businesses.

Analyzing Data Retention Policies

Despite the move toward greater transparency in training preferences, Google has simultaneously adjusted its data retention policy. Temporary chats—those conducted with history turned off—will now be retained by Google for up to two months, a significant increase from the previous 72-hour window. The company justifies this extension as a necessary measure to improve safety and security standards as AI models become more complex and capable of performing autonomous actions.

Competitive Context and Industry Standards

This policy change places Google in a unique position relative to its competitors. While OpenAI and Anthropic generally retain temporary chats for approximately 30 days, Google’s new two-month window is notably longer. This creates a complex landscape where users must weigh the benefits of advanced, agentic AI features—like those found in 'Tasks'—against the reality of longer data retention periods, even when privacy settings are ostensibly hardened.

Broader Implications for AI Privacy

These developments highlight a broader trend in the tech industry: the tension between 'agentic' AI functionality and user privacy. As AI evolves from simple chatbots into agents capable of making purchases and calling businesses, the amount of sensitive data processed by these models increases exponentially. Google’s dual approach of providing more granular training opt-outs while extending retention windows suggests a move toward balancing legal compliance and safety protocols with the technical requirements of building more reliable AI assistants.

Future Trends and Conclusion

Looking ahead, the success of these updates will likely depend on how clearly Google communicates these changes to its global user base. As the 'Tasks' feature rolls out to more regions, the ability to maintain privacy without sacrificing utility will become a primary differentiator in the competitive AI market. Ultimately, these updates represent a shift toward a more integrated, yet potentially more data-intensive, user experience.

Verification Required?

Read the full report from the primary source

Go to Ars Technica