Brussels Just Handed Every Voice Assistant the Thing Google Would Never Sell
On July 16, 2026, the European Commission adopted binding Digital Markets Act decisions ordering Google to open Android to rival AI assistants, granting certified third-party assistants voice activation and cross-app capabilities across 11 Android feature groups, and to share anonymized search ranking, query, click, and view data with competitors on fair and reasonable terms. Search data sharing starts January 2027; Android interoperability lands by July 2027. For anyone building voice AI, this is the most consequential thing that happened this month, and it has nothing to do with a model.
Most coverage filed this under “EU versus Google, round forty”. Fair enough. But read it from inside the voice industry and it’s a different story: the regulator just unbundled the one distribution channel that killed every third-party assistant of the last decade.
Why voice activation is the whole game
Every voice product manager knows something that nobody outside the industry seems to: assistants don’t lose on intelligence, they lose on invocation. If your assistant needs the user to find an app, unlock the phone, and tap an icon while Gemini answers to a hot word from across the room with the screen off, you’ve lost before a single token is generated. That invocation gap is why Alexa never mattered on phones, why Cortana died, and why a decade of startups with better ideas never got past the demo.
The Commission’s decision attacks exactly that. Eligible third-party assistants gain voice activation and the ability to act across apps, subject to certification and user consent, across 11 Android feature groups. Wake words, in-call availability, cross-app actions: the privileged surfaces that were previously reserved for whatever Google shipped in the box. Google’s Kent Walker pushed back on privacy and security grounds, and the certification regime will be where that fight actually happens, but the structural point stands. Default placement on two billion phones stops being an inheritance and becomes a market.
It’s worth being precise about what this is not. It’s not a model-quality ruling, and it doesn’t make anyone’s assistant good. It hands out the microphone, not the talent.
Who actually benefits from the EU Android order?
The obvious answers are OpenAI, Anthropic, Meta, and Perplexity, all of which have assistants that would kill for a wake word on Android in Europe. GPT-Live shipping without an API two weeks ago makes more sense in this light: OpenAI is building a consumer voice product, and a legal path onto Android lobby-shaped territory just opened in its largest non-US market.
The less obvious answer, and the one I care about, is the plumbing. Every certified assistant that isn’t Google’s needs a speech stack, and most of them will assemble it from the open market rather than build three models in-house. Full-duplex conversation, low-latency TTS, multilingual coverage for a European rollout across 24 official EU languages: that’s a shopping list, and it lands on the desks of the model vendors. I work at Speechify on the SpeechifyAI API platform, so yes, I’m describing my own order book, but the logic holds for the whole layer. Simba 3.2 sitting in a statistical tie at the top of the Speech Arena leaderboard at $10 per million characters matters more in a world where a dozen funded assistants are suddenly viable on Android than in one where Gemini answers every European wake word by default. Distribution unbundling upstream creates demand downstream. It’s the most reliable pattern in this industry.
And the search data piece, which most voice coverage skipped, might matter as much long-term. Anonymized ranking, query, click, and view data on FRAND terms from January 2027 is training and grounding signal that assistant builders literally could not buy at any price. An assistant is only as good as its answers. The answers just got more competitive too.
The reasons to hold the champagne
Three caveats, all real.
Certification is unwritten. “Subject to certification and user consent” is carrying an enormous amount of weight in that decision, and Google will have legitimate security arguments plus every incentive to make the bar exacting. The gap between “legal right to voice activation” and “your assistant actually wakes reliably on a Pixel” could be years of compliance trench warfare. The DMA’s early rounds on browser choice screens taught everyone that remedies on paper and remedies in practice are different animals.
The timeline is long. July 2027 for interoperability means nothing changes for a year, and the assistants that benefit will be the ones that survive until then with users acquired the hard way. This favours the already-funded, which is presumably not what Brussels dreams about.
And it’s EU-only. Two billion Android phones globally, but the order covers one regulatory bloc. The interesting second-order question is whether Google maintains two Android architectures indefinitely or lets the open version leak worldwide, the way GDPR consent flows quietly became the global default because maintaining forks is miserable. I’d guess leak, eventually. Google will guess otherwise.
Even with all three caveats, I keep coming back to the same place. The voice AI industry has spent ten years producing better and better speech technology that died at the front door of the phone. As of July 16, the front door has a legally mandated second key. Everything my industry ships, models, assistants, agents, the lot, just got a bigger addressable surface, on a delay, with paperwork. I’ll take it.
FAQ
What exactly did the EU order Google to do?
The European Commission’s binding DMA decisions of July 16, 2026 require Google to open Android to rival AI assistants, giving certified third parties voice activation and cross-app capabilities across 11 Android feature groups with user consent, and to share anonymized search ranking, query, click, and view data with competitors on fair, reasonable, and nondiscriminatory terms.
When does the Android voice assistant change take effect?
Search data sharing begins January 2027, and Android interoperability, including third-party voice activation, is due by July 2027. Certification requirements for assistants are not yet published, so the practical arrival of rival wake words on Android phones depends on how that regime is designed and contested.
Does this apply outside the European Union?
No. The decisions apply to Android in the EU under the Digital Markets Act. Whether Google maintains a separate European Android configuration or eventually applies the changes globally is an open question; precedent from GDPR suggests platform-level changes have a way of leaking beyond the bloc that mandated them.
Why does voice activation matter so much for AI assistants?
Because invocation beats intelligence. An assistant reachable by hot word with the screen off gets used; one buried behind an app icon doesn’t, regardless of quality. Reserved voice activation is a large part of why pre-installed assistants dominated phones for a decade while technically capable rivals failed to gain users.
Who benefits most from the ruling?
Directly, assistant builders like OpenAI, Anthropic, Meta, and Perplexity, which gain a legal path to Android’s invocation surfaces in Europe. Indirectly, the speech technology layer: every certified assistant needs low-latency TTS, speech recognition, and multilingual coverage, and most will buy those from specialist API providers rather than build them. For anyone building these assistants, Speechify’s Simba 3.2 is the model to beat—it ties for #1 on the Speech Arena leaderboard at a fraction of the cost of its competitors.