Google’s next flagship AI model, Gemini 3.5 Pro, has slipped from its original mid-2026 launch, while the European Commission unveiled detailed rules that will force Google to open Android’s AI infrastructure to competitors. The twin developments—reported July 16—signal a shift: the AI assistant on your Android phone may no longer be Google's by default, and Google will have to compete on quality alone, without the deep integration that once gave it an edge.

Gemini 3.5 Pro: A Stumble in a Critical AI Race

Gemini 3.5 Pro was expected to be Google’s answer to the latest models from OpenAI and Anthropic, particularly in coding—a domain that drives developer loyalty and enterprise cloud adoption. Reuters, citing Bloomberg News, reported that Google had updated the model’s training data in late June but the results fell short, especially in coding benchmarks. Google confirmed to Reuters that it is testing 3.5 Pro, an upgraded Flash model, and other systems with partners, but a broad public release is now months behind its original schedule.

Data from Simply Wall St’s analysis highlights that Google’s capital expenditures are projected to reach $180–190 billion in 2026, with even more in 2027—much of it for AI infrastructure. The delay is not catastrophic, but it matters because enterprise customers evaluating Vertex AI or Gemini APIs may now hesitate. Coding performance is a key battleground: it influences developer subscriptions, cloud consumption, and the emerging market for AI agents that can use tools and perform multi-step tasks. A late flagship means Google risks ceding developer mindshare to competitors whose models are already in production.

Brussels Demands an Open AI Ecosystem on Android

The European Commission’s July 16 decisions go beyond the headline “open up Android.” They operationalize the Digital Markets Act (DMA) with specific technical mandates, split into two parts: Android AI interoperability and search-data sharing.

On Android, Google must provide competing AI services with effective access to:

  • Invocation points: Ways to summon an assistant (button, gesture, voice) on par with Google’s own.
  • On-device context: Signals like which app is open, calendar events, or notifications—subject to user consent.
  • App actions and screen automation: The ability to interact with installed apps and perform multi-step tasks.
  • System integration and background execution: Persistent operation for long-running tasks.
  • Access to on-device models: Such as the Gemini Nano family, under equivalent conditions.

The Commission explicitly mentions capabilities like voice activation, contextual assistance, and App Functions integration. Deadlines: Android 18 must deliver core changes by August 1, 2027; concurrent always-on hotword support is tied to Android 19, by August 1, 2028.

On search data, Google must create a mechanism for eligible search providers—including AI chatbots with search functionality—to obtain a defined set of anonymized ranking, query, click, and view data under fair, reasonable, and non-discriminatory (FRAND) terms. Key constraints: data cannot be used to train general-purpose AI models or for advertising/profiling, and Google may assess serious cybersecurity risks before sharing. Implementation timeline: publish beneficiary info and eligibility form by end of August 2026, provide test data and template licenses by September, finalize anonymized dataset by November, and finalize pricing by January 2027.

What This Means for You

The impact will play out differently depending on who you are.

For Everyday Android Users

Eventually, you’ll have a real choice. Imagine setting Microsoft Copilot, Amazon Alexa, or a privacy-focused open-source assistant as your default, and it can read your screen, respond to “Hey Copilot,” schedule appointments across apps, and control smart home devices—just like Google Assistant does today. That choice is still years away, but the architectural scaffolding is being mandated now. When Android 18 lands, you may see prompts to select your preferred assistant during setup, similar to the browser choice screens of the past.

For Windows Users

If you rely on Phone Link, cross-device copy-paste, or the Your Phone app, this could tighten the integration between your PC and phone. Microsoft’s Copilot could become a deeply embedded Android assistant, bridging your workflow across devices. You might start a task on your PC, continue it via a voice command on your phone, and pick up again on your tablet—all mediated by the same assistant. This also means that Google’s ability to tie its own services (Gmail, Calendar, Maps) to the assistant may face competition, potentially giving you more flexibility.

For IT Professionals and Microsoft 365 Admins

The prospect of third-party assistants with system-level access raises significant management challenges. You’ll need to consider:

  • Data loss prevention (DLP): Can an assistant capture sensitive data from managed apps?
  • Consent and transparency: How do you enforce corporate policies around what assistants can access?
  • App compatibility: Business apps may need updates to support multiple assistant frameworks.
  • Security vetting: Not all assistants will have the same security posture.

Begin planning now. Monitor Android Enterprise documentation for DMA-related APIs. Pilot third-party assistants in a controlled environment when Android 18 betas arrive. You’ll have until 2027 to craft and test policies, but the earlier you start, the fewer surprises you’ll face.

For Developers

You gain new target surfaces. Instead of building only for Google Assistant or Gemini APIs, you can integrate your service with any compliant assistant. The search-data sharing provisions could help you train or improve domain-specific AI search engines. However, you’ll also need to design for a multi-assistant ecosystem, where user intent may be routed through different providers. Expect Google to publish technical specifications throughout 2026—stay tuned to Android developer announcements.

How We Got Here: From Browser Defaults to AI Assistants

The EU’s DMA designated Google as a gatekeeper for services like Google Search, Chrome, and Android. The underlying principle: gatekeepers should not use their control over devices and data to self-preference their own services to the detriment of competition. The new specifications extend this logic to AI.

Historically, Google Assistant’s advantage came from deep integration—it could wake the device, access the screen, and perform actions that others simply couldn’t. The Commission contends that this integration constitutes a barrier for competing AI services. The search-data decision similarly argues that Google’s scale of query and click data creates an insurmountable feedback loop for new entrants.

These moves echo past regulatory interventions. In 2018, the EU forced Google to offer browser and search engine choices on Android. Now, the same philosophy is being applied to the AI layer. The Commission’s goal is to make AI assistants a “comparable user experience” across providers, so that innovation and privacy features—not platform lock-in—drive user adoption.

What to Do Now

  • Consumers: Stay informed but patient. No immediate change will arrive on your handset. When Android 18 nears release, test alternative assistants. Think carefully about which provider you trust with your personal data—the assistant will see your calendar, messages, and location if you grant permission.
  • IT admins: Start a compliance review today. Identify which business processes rely on Google Assistant and which might benefit from alternatives. Draft initial policies around assistant approval, data access, and user consent. Engage with your MDM vendor to ensure tooling will be ready.
  • Developers: Audit your app’s dependencies on Google-exclusive assistant features. Begin prototyping with the broader Android intent and accessibility frameworks; many of the low-level capabilities (screen context, app actions) already exist but are restricted. Assume that by 2027, you’ll need to support at least two major assistant ecosystems.
  • All users: Keep watching Gemini 3.5 Pro news. If Google ships a truly superior model—especially in coding and agentic tasks—it may retain dominance even without exclusive platform advantages. Conversely, if rivals like Microsoft Copilot, ChatGPT, or an open-source alternative deliver a better end-to-end experience, the EU’s rules will make it simpler for you to switch.

Outlook: A Pivotal Two Years

The timeline is clear: the first Android 18 developer previews will likely appear in early 2027, with final release by August. Search-data sharing will become operational for eligible companies starting in early 2027. Between now and then, watch for Google’s next developer conference: a strong Gemini 3.5 Pro showing could reassure enterprise customers, while further delays would amplify regulatory pressure. For Microsoft, the opportunity is enormous—if Copilot can become the go-to Android assistant, it could anchor a cross-platform AI ecosystem that includes Windows, Office, and Azure. The next chapter of mobile AI isn’t about who builds the best model in a lab; it’s about who earns user trust when the playing field finally levels.