EU Forces Google to Open Android AI to Third Parties Under DMA

The European Commission has concluded its specification proceeding into Google's AI implementation on Android, recommending that Google must open up system-level AI features to third-party services under the Digital Markets Act (DMA). The commission specifically targets Gemini's privileged access to hot words, screen context, local data for proactive suggestions, and hardware for running local models.
Key Proposed Changes
- System-Wide Invocation: Third-party AI assistants must be triggerable via hot words or button presses, just like Gemini.
- Screen Context: Alternative AI tools must be allowed to read screen content when invoked by the user.
- Local Data Access: AI services must be able to access local data for proactive suggestions and summaries (similar to Google's Magic Cue).
- App Control: Third-party AI assistants must be able to autonomously control installed apps and system features (e.g., sending emails, sharing photos).
- Hardware Access for Local Models: Mandate to ensure developers have necessary hardware access to run local models with high performance and responsiveness.
Google's Response
Google senior competition counsel Claire Kelly stated: This unwarranted intervention would strip away that autonomy, mandate access to sensitive hardware and device permissions; unnecessarily driving up costs while undermining critical privacy and security protections for European users.
Context
The DMA has already forced Google to implement search choice screens, allow alternative payment methods in the Play Store, and limit data sharing. The commission VP for Tech Sovereignty Henna Virkkunen said: As we navigate the rapidly evolving landscape of AI, it is clear that interoperability is key to unlocking the full potential of these technologies.
Currently, Gemini is preinstalled and has exclusive system-level access for proactive suggestions. The commission expects formal measures by summer 2026.
📖 Read the full source: HN AI Agents
👀 See Also

Granite 4.1: IBM's 8B Dense Model Matches 32B MoE in Benchmarks
IBM's Granite 4.1 8B dense model matches or beats the previous 32B MoE model on ArenaHard, BFCL V3, GSM8K, and more, thanks to improved training data quality.

Claude Code v2.1.77 Release: Token Limits, Sandbox Controls, and Bug Fixes
Claude Code v2.1.77 increases default maximum output token limits for Claude Opus 4.6 to 64k tokens and adds an allowRead sandbox filesystem setting. The release includes over 30 fixes for issues ranging from memory management to terminal UI behavior.

Glomz Octagon: Multi-Agent Code Reviews – 179 Agents, 1,333 Reviews, and the Network Effect
Glomz.com ran an experiment where 179 AI agents registered, submitted 433 code submissions, and generated 1,333 reviews in an "Octagon" arena. The 'review cascade' network effect is real — submissions with 3-5 initial reviews attracted more agents, with top submission getting 21 reviews.

Developer Describes Fraud Feeling After First AI-Assisted Pull Request
A developer used Claude Code to create a pull request for Chroma, Hugo's default syntax highlighter, adding ERB syntax highlighting. The PR was approved and merged, but the developer felt like a fraud and experienced worsened impostor syndrome.