EU Forces Google to Open Android Wake Word and Assistant APIs to All

The European Commission has adopted a Digital Markets Act (DMA) decision requiring Alphabet to open up eleven Android features to third-party assistants on equal terms. The ruling, announced July 16, 2026, directly targets Google's stranglehold on wake word detection, which had forced Home Assistant to run a CPU-based workaround that drained batteries and kept the microphone privacy indicator permanently on.

For three years, the Open Home Foundation tried to ship always-on wake word detection in its Android Companion app. The blocker: Android's efficient two-stage wake word architecture—first-stage detection on a low-power DSP, second-stage confirmation on the CPU—was only accessible to Google and device manufacturers. Third-party apps had to run the entire detection on the CPU, jumping battery usage from roughly 1% to 15% and requiring full microphone access, which tripped the privacy indicator and forced users to trust the app with raw audio.

The DMA decision requires Google to provide "equally effective" interoperability across eleven features, including wake word detection, ambient sensor access, and screen automation. Key provisions for wake words: third parties can create custom wake word models with first-stage detection on the DSP, run second-stage validation, and get testing tools and documentation without a commercial agreement. The decision also decouples feature access from default assistant roles and mandates concurrent wake word detection from multiple services—so users can say "Okay Nabu" to control their home while still using Gemini for other tasks.

Google must ship these changes in Android 18 by August 1, 2027, with concurrent hotword detection in Android 19 by August 1, 2028. The Commission included safeguards against malicious compliance: solutions must be "equally effective" on ease of use, speed, and energy consumption, with complete documentation and monthly progress reports.

For Home Assistant, this unlocks battery-efficient wake word detection on DSP-equipped phones, the ability to run multiple assistants side-by-side, and OS-enforced sandboxing that guarantees audio isn't sent anywhere until the wake word is confirmed—privacy by design rather than by trust.