Google bets Gemini's future on data you may not want to share

The News

Google used its I/O 2026 keynote on Tuesday to weave artificial intelligence deeper across its consumer products, with one common requirement: deep access to a user's personal data. The centrepiece was Gemini Spark, billed as an always-on AI agent that can organise upcoming events and act on a user's behalf inside Google's apps.

Spark sits alongside Daily Brief, a feature that hands users a personalised rundown of what to expect from their day, drawing on calendar entries, location signals and inbox content. Google also expanded access to Gmail's AI inbox, which can now build custom to-do lists and draft replies in a tone modelled on past emails, a fuller version of features previously gated to Workspace Labs participants.

The company unveiled a new Gemini 3.5 model family, fresh Search and Gmail integrations and an update to Project Aura, its smart-glasses programme, but the through-line of the keynote was that every benefit depends on letting Gemini read more of you.

Why It Matters

The trade is not new, but the scale is. The last time a personal assistant rollout depended this heavily on inbox and calendar access was Google's first push for Gmail's smart compose in 2018, which still drew years of regulatory pushback in the EU. Spark and Daily Brief go further: they are persistent rather than reactive, and they aim to summarise and act on data the user has not specifically asked Google to touch.

That changes the consent calculus. A one-off feature like an inbox summariser is easy to switch on or off. A persistent agent that quietly stitches together mail, search history, location and contacts is closer to a profile than a tool. Investors will read this as Google trying to lock users into a habit loop that competitors such as OpenAI's ChatGPT and Anthropic's Claude have struggled to build because they lack the underlying data graph.

Trust, in other words, is now Gemini's moat, and Google's biggest risk if it slips.

Indian Angle

For India, the moat sits squarely inside the Digital Personal Data Protection Act 2023, whose implementing rules are being phased in through MeitY notifications. The DPDPA requires specific, informed consent for each processing purpose, restricts the use of children's data, and forces data fiduciaries to honour withdrawal of consent. An always-on agent that reads mail to draft replies and scans calendars to brief the user looks like several distinct processing purposes bundled into one switch, exactly the bundled-consent pattern the law was written to constrain.

There is also a competition angle. Indian model builders such as Sarvam and Krutrim have positioned themselves as alternatives that can be trained and deployed on Indian infrastructure under MeitY's IndiaAI compute commitments. If Google ties Spark deeply into Workspace, which dominates Indian enterprise and education, those domestic players will find it harder to win the consumer mind-share that tends to follow enterprise habits.

For Indian developers and start-ups, the more immediate question is pricing. Gemini 3.5's API rates will set the floor for any India-built copilot that competes on quality, and a sharp price drop, of the kind Google has used at past I/Os to undercut OpenAI, would compress margins for every domestic AI wrapper currently billing in rupees.

FAQ

When do Spark and Daily Brief reach Indian users?

Google has not confirmed an India-specific launch date. Past I/O features such as the original Bard rollout reached India four to six weeks after US availability, and Spark is expected to follow a similar pattern, gated by DPDPA consent screens once MeitY's operational rules take full effect across the country.

How does Gemini 3.5 compare to the previous version?

The Verge's roundup describes 3.5 as a new family rather than a single model, suggesting tiered variants for speed and capability, similar to the layout used for Gemini 2.0 and 2.5. Specific benchmark figures and India pricing were not announced on stage at the keynote.

What does this mean for Indian businesses using Gmail and Workspace?

Indian organisations will need to update their data-protection notices and revisit vendor due diligence under the DPDPA before enabling AI inbox features for staff, especially teams handling sensitive customer data such as bank statements, PAN details or health records. Bundled-consent flows should be reviewed first.

Where can I read the original announcement?

The Verge has the fullest day-one breakdown of the keynote, including the trust and data-access angle covered here. Google's own blog post provides the official feature list, country-by-country rollout timing, and the privacy controls Google says users will receive for each new agent.

This story was reported by The Verge. Read the full original coverage at The Verge.