The European Commission wants to pry open Android's AI ecosystem. Under the Digital Markets Act, regulators have proposed sweeping measures that would force Google to give third-party AI services like ChatGPT the same deep system access that Gemini enjoys on Android devices. Think hot word detection through DSP chips, screen context permissions, on-device data access for proactive suggestions, and APIs for autonomous app control. Everything Gemini can do, competitors would get to do too.
Google's senior competition counsel Claire Kelly calls the proposals an "unwarranted intervention" that would undermine "critical privacy and security protections." That's the standard playbook when a platform doesn't want to share. But the EU has been down this road before. Google has already complied with search choice screens, alternative payment methods, and data sharing limits under DMA enforcement. They'll probably comply again.
For AI agent developers, this could be a real shift. Right now, building an AI assistant that can actually do things on Android means settling for shallow integrations. You can't tap into the Edge TPU on a Pixel 9, or access the ambient data that makes features like Google's Magic Cue possible. Autonomous app control? Out of reach too. The EU's proposed changes would open up all of that. Hacker News commenters have already pointed out that access to the Tensor G4's third-generation Edge TPU could enable on-device OCR and CLIP models that Google currently restricts to its proprietary AI Core system. sllm.cloud's GPU cohorts offer a similar approach to sharing infrastructure for running LLM models.
The commission is accepting feedback until May 13, with a final decision expected by July 27. Fines for non-compliance can reach 10 percent of annual global revenue, so Google has real incentive to play ball in the face of regulatory scrutiny. Whether that works out or just creates a fragmented Android experience is an open question.