The EU ordered Google to give ChatGPT and Claude the same Android access Gemini has

The European Commission adopted a binding decision on July 16, 2026 ordering Google to give competing AI assistants — including OpenAI's ChatGPT, Anthropic's Claude, and Perplexity — the same system-level access to Android that Google's own Gemini currently holds exclusively. The Commission's own stated rationale: roughly 60% of EU Android users currently have no equal-footing AI alternative to Gemini, because only Gemini gets full access to the operating system's voice, context, and action layers.

This wasn't a snap decision. The Commission opened two parallel specification proceedings against Google back on January 27, 2026 — case DMA.100220, under Article 6(7) of the Digital Markets Act — each carrying a six-month statutory deadline that the July 16 decision met almost to the day. That timeline matters because it means the substance was worked out over six months of proceedings against Google's own gatekeeper obligations under the DMA, not announced cold: Google has had the specific 11 Android features at issue on the table since the proceedings opened, and the July 16 decision is the codified outcome of that process rather than a new surprise obligation.

What Google actually has to open up

The decision spans 11 specific Android features, grouped into four capability layers. Invocation: rival assistants must be reachable the same way Gemini is — by voice command or a system-level access point like the home button, not buried behind extra taps. Actions: rival assistants must be able to complete tasks in and across apps, including long-running ones, not just answer questions. Context: rival assistants get access to on-screen and device-sensor context so they can act proactively instead of only when asked. Resources: Google has to make its own on-device AI models and adequate hardware/software resources available to rival assistants, rather than reserving the best-performing path for Gemini. Privacy, security, and device-integrity protections remain in place across all four.

A second, separate specification requires Google to share anonymized Search data — query, click, and ranking signals — with competing search engines and AI chatbots that perform equivalent retrieval, under fair, reasonable, and non-discriminatory pricing, starting January 2027. The Android interoperability changes above are due by August 1, 2027 — some reporting rounds this to "July 2027."

What happens if Google doesn't comply

Non-compliance with the specification decision itself risks fines of up to 10% of Alphabet's annual worldwide revenue — a figure that would run into the tens of billions of dollars at Alphabet's current scale. That penalty is separate from, and larger in scale than, the €890 million ($1.01 billion) fine the Commission issued a week later, on July 23, for different DMA violations: €460 million over Google favoring its own services in Search rankings, and €430 million over restricting how app developers can steer users to non-Play-Store payment options. Google was given 60 days to fix those specific practices or face further periodic penalties of up to 5% of worldwide turnover. The two cases share a law and a defendant, not a subject.

Concretely, the gap being closed is this: a Gemini competitor on Android today can typically answer a question, but it can't be invoked the same way Gemini can (by a home-button long-press or a system-wide voice trigger), can't read on-screen content to act on it automatically, and can't call on Android's own on-device AI models the way Google's assistant does. The July 16 decision requires Google to close those specific gaps by the deadlines above — not simply to publish a developer API that rivals could theoretically use but Google could leave impractical in practice.

Google had not issued a public statement specifically addressing the July 16 interoperability decision as of publication. In past EU proceedings over similar Android and DMA obligations, Google has argued that mandated interoperability threatens Android's security model and discourages investment in the platform — but that record predates this specific decision and shouldn't be read as its response to it.

This is the same Digital Markets Act machinery that has spent 2026 pushing into AI specifically — distinct from, but adjacent to, the export-control and national-security lens US regulators have applied to [Chinese humanoid robots](#/article/fcc-covered-list-bans-chinese-humanoid-robots) this year. Where Washington's actions this year have mostly restricted market access on security grounds, Brussels' DMA actions are aimed at forcing access open on competition grounds — opposite instincts, same underlying fact that AI assistants are now infrastructure-level products regulators treat as chokepoints. [OpenAI](#/company/openai) and [Anthropic](#/company/anthropic) are named beneficiaries here without having sought anything from the EU directly — the remedy runs through Google's own platform obligations, not through them.

Scope matters here as much as substance: the obligations bind Google's Android build within the EU specifically, enforced directly by the Commission rather than by any national regulator — which is what let two proceedings opened against Google in January converge into one decision applying uniformly across all 27 member states rather than fragmenting into parallel national cases. Nothing in this decision changes what Android looks like, or what rival assistants can do, outside the bloc.