Version 1

At Google I/O 2026, Google officially declared that 'Google search is AI search,' signaling a fundamental identity shift in its core product [1]. AI Mode now reaches over 1 billion users per month, doubling quarter-over-quarter since launch [1]. Google simultaneously announced Gemini 3.5 Flash, an agent-optimized model claimed to surpass the prior-generation Pro model in intelligence while remaining efficient enough for complex agentic tasks at scale [2]. Alongside these, Google unveiled Gemini Spark, a personal AI agent integrating natively with Gmail, Calendar, Drive, and other Google services, built on the closed-source Antigravity platform [3].

Google is reshaping search and personal computing for over a billion users simultaneously — and because it controls the platform, user objections are unlikely to slow the transformation [1]. The Gemini Spark agent's access to deeply personal data raises acute security concerns, particularly around prompt injection, a vulnerability Google has not explicitly addressed in its stated security posture [3].

Google I/O 2026 marked what the company positioned as an inflection point for its core products. Google Search VP Liz Reid declared 'Google search is AI search,' framing the shift not as a feature addition but as a redefinition of the product's identity [1]. AI Mode — Google's conversational search layer — now reaches over 1 billion users per month, having doubled every quarter since its launch, and remains free for all Google search users unlike most of the company's other AI products [1]. Ars Technica's Ryan Whitwam observed that Google's sheer market scale means it can drive this transformation regardless of user preference, with its own engagement metrics serving as the sole validation — a dynamic that forecloses meaningful user resistance [1].

On the model side, Google announced Gemini 3.5 Flash as an agent-optimized model designed to bring frontier-level intelligence to complex, multi-step tasks at practical cost and speed [2]. Google claims it outperforms the previous generation's Pro model — a pattern that has repeated with each tick-tock update cycle — though Whitwam noted cautious skepticism about whether this cycle's claims are meaningfully different from prior ones [2]. Gemini 3.5 Flash is rolling out immediately across a wide range of Google products and serves as the model powering Gemini Spark [2][3].

Gemini Spark is Google's new personal AI agent, announced at I/O and built on the Antigravity platform — a closed-source Go binary with an open-source Python SDK wrapper [3]. It connects natively with Gmail, Calendar, Drive, Docs, Sheets, YouTube, and Google Maps, and is positioned for both consumer and enterprise use. Google cited ephemeral VMs and DLP policies as its enterprise security foundation, but developer and security commentator Simon Willison noted that neither measure specifically addresses prompt injection — the attack vector by which a malicious email or document could hijack the agent's behavior to exfiltrate or manipulate other data [3]. Willison called Gemini Spark a 'top candidate for the agent security challenger disaster that we still haven't seen,' given the sensitivity of the data the agent will routinely process [3].

A secondary concern raised by Willison is Google's decision to deprecate the Apache 2.0-licensed Gemini CLI and cut off its AI subscription compatibility on June 18th, replacing it with the closed-source Antigravity CLI [3]. This move signals Google's intent to consolidate its agentic infrastructure under proprietary tooling, a strategic choice that has drawn criticism from developers who had built on the open-licensed toolchain.