Google Messages Introduces Tap to Draft for Safer, Editable Smart Replies and Enhanced RCS Security with Airtel in India

Digital messaging is quietly entering a new governance phase, where AI-generated text and telecom-grade security controls are treated less as convenience features and more as infrastructure that has to be auditable, user‑controllable, and compliant with national rules on spam, privacy, and consent.

Google is reshaping how AI-assisted texting behaves in Google Messages with a user-controlled option that places suggested replies into the compose box before anything is sent. The feature – “Tap to Draft” – is now surfacing to beta users, giving people a chance to review and personalize Smart Replies rather than firing them off instantly.

Smart Replies gain a review step, reducing misfires

Until now, Smart Replies favored speed with a “Tap to Send” model. The new path inserts a deliberate pause: suggestions land as editable text you can tweak or discard. That small UX change lowers the risk of accidental or out‑of‑tone responses while keeping the convenience of one‑tap assistance. It also effectively makes the user a final editor of AI output rather than a passive sender, a distinction regulators and workplace compliance teams increasingly care about as AI-generated content appears in more professional and cross‑border exchanges.

Availability and how to enable it

• Feature channel: Google Messages beta.
• Observed build: 20260303_00_RC00.
• Path: Settings → Suggestions, then toggle between “Tap to Send” and “Tap to Draft”.
• Status: Not yet in the stable app; broader availability is expected in the coming weeks.

Users who want early access can join the Google Messages beta and check for updates. For organizations that manage fleets of Android devices, that rollout path also determines how quickly corporate messaging policies can take advantage of the additional review step.

The change does not alter existing end‑to‑end encryption for personal RCS chats; it only adjusts how suggested text is staged before you press send. That distinction matters for compliance teams: content remains protected in transit for eligible chats, while the AI assistance layer remains a client-side aid rather than a new data‑sharing channel.

Why the extra tap matters for AI in everyday chat

Smart Replies are generated by on‑device and cloud‑assisted models that infer likely responses from conversational context. Inserting a draft step builds friction where it matters: users can correct names, add nuance, or switch tone without battling the clock. That makes AI assistance feel less mechanical and more accountable, particularly in work or cross‑cultural conversations where phrasing carries weight and misinterpretation can have legal, HR, or diplomatic consequences.

For public institutions and regulated industries experimenting with AI in communications, the pattern is instructive: automation proposes, a human disposes. That “human‑in‑the‑loop” approach is increasingly echoed in draft AI governance frameworks and internal policies that require a clear moment of human review before machine‑generated text leaves an official channel.

Airtel and Google team up to tighten RCS security in India

Alongside the usability change, Google and Airtel are collaborating to offer an AI‑powered, safer Rich Communications Services experience across Airtel’s network in India. The initiative blends RCS features like high‑quality photo/video and message reactions with telecom‑grade checks intended to curb spam and fraud, in a market where mobile messaging is a primary channel for banking, public‑service alerts, and political communication.

• Real‑time security checks verify sender identity using telecom‑based business verification systems.
• Do Not Disturb (DND) preferences are honored by classifying traffic as promotional or transactional and applying limits where required.
• Spam business messages are stopped before delivery, with harmful links blocked through multiple layers of threat detection.
• Messages from senders flagged by Google and Airtel’s AI‑based spam detection tools are rate‑limited or suppressed.

These controls align with India’s enterprise messaging rules, where traffic types are distinguished and customer preferences must be enforced under the Telecom Commercial Communications Customer Preference Regulations administered by the national regulator, the Telecom Regulatory Authority of India. Users can also manage their marketing opt‑outs via national DND preferences, which function as a statutory backstop for consent and redress when spam filters fail.

System design: how protections layer on top of RCS

• Transport: RCS uses IP data rather than legacy SMS/MMS. Google’s cloud infrastructure brokers interoperability and business messaging with verified sender identities, giving carriers and regulators clearer levers to distinguish legitimate enterprises from spoofed or grey‑route traffic.
• Personal chats: One‑to‑one and group RCS conversations in Google Messages support end‑to‑end encryption; SMS/MMS are not encrypted.
• Business traffic: Brand verification, traffic classification, URL risk checks, rate limiting, and user reporting provide defense‑in‑depth against scams; business RCS is authenticated but not end‑to‑end encrypted, reflecting a trade‑off between security oversight and privacy that policymakers are still actively debating.
• User controls: Report and block functions feed model‑driven spam detection; per‑thread settings let users mute or restrict attention‑grabbing behaviors, creating a feedback loop that both platforms and regulators can point to when assessing whether user consent is meaningful in practice.

Quick setup tips and safety checklist

• Update to the latest beta build, then open Google Messages → Settings → Suggestions and choose “Tap to Draft”.
• Use the draft to adjust formality, fix names, and add context before sending, especially in professional or sensitive threads.
• Disable Smart Reply entirely if you prefer full‑manual composition or your organization requires it for certain roles.
• Treat unexpected links cautiously. Prefer interacting only with verified business senders and known institutions.
• In India, review or update your DND category settings to limit unwanted marketing outreach and ensure your registered preferences match the reality of what you receive.

By pairing a human‑in‑the‑loop “Tap to Draft” experience with carrier‑level controls on RCS, Google is nudging messaging toward a model where automation accelerates routine tasks while people retain editorial judgment and security remains a first‑order design constraint. For regulators, enterprises, and public bodies, these design decisions are becoming reference points in a broader argument: AI can sit inside mainstream communications, but only if people – and the rules that protect them – stay firmly in the loop.