EU Orders Google to Share Search Data and Open Android to AI Competition

On July 16, 2026, the EU formally adopted regulations requiring Google to **share search data** and **open up Android to AI competition**. Google will be obligated to provide competitors with core search assets such as link targets, query data, and user behavior data. The company is also required to open its platform so that providers other than Google can offer AI assistants and agents on Android devices on equal terms.

The move aims to foster a competitive environment for the AI era, premised on Google's dominant position in both the search and mobile OS markets. It is grounded in the European Commission's finding that unequal access to data is hindering competition as AI-powered search gains traction. Google will now be pressed to respond on the details of the remedies and the compliance schedule.

As a step toward concrete enforcement of the Digital Markets Act (DMA), the decision could become a global precedent for AI-related platform regulation.

AWS Bedrock Managed Knowledge Base Reaches GA — Fully Managed Enterprise RAG

AWS has announced the **general availability (GA)** of Amazon Bedrock Managed Knowledge Base. The service provides a retrieval foundation (RAG) for agents that connect to enterprise data — without having to build connectors, parsers, vector stores, or retrieval logic separately.

Key features include:

Six native connectors**: S3, SharePoint, Confluence, Google Drive, OneDrive, and Web Crawler, with incremental sync to keep costs down

Real-time ACL checks**: Permissions are verified directly against the source of truth at query time, ensuring the LLM never sees unauthorized data

Multimodal parsing**: Automatic analysis of PDF, PPTX, DOCX, audio (up to 2GB), and video (up to 10GB)

Agentic Retrieval**: Decomposes complex queries into subqueries and performs iterative retrieval, repeating planning and evaluation for up to five rounds

AgentCore Gateway integration**: MCP-compatible agents can automatically discover and invoke the knowledge base as a tool

OpenAI, Sony, Syngenta, and others have publicly shared their early adoption. The service is available in eight regions, including Tokyo (ap-northeast-1). Pricing is usage-based for storage, the standard retrieval API, and Agentic Retrieval, with the multimodal parser, embedding models, and reranker included at no additional cost.

Netflix Discloses Generative AI Use Across Around 300 Titles

In its Q2 2026 earnings report, Netflix revealed that it has used generative AI across **around 300 titles**. While no detailed breakdown of specific applications (video generation, script assistance, localization, VFX, etc.) was provided, it is symbolically significant that a major streaming service disclosed the scale of its AI usage in a public quarterly earnings setting.

As one indicator of how deeply generative AI has penetrated the entertainment industry, attention will now focus on whether other major platforms follow suit and how creator organizations respond. As of today, Netflix has offered only limited additional comment, so the caveat stands that the impact on production workflows "remains impossible to determine."

Microsoft 365 Copilot Struggles, with Business Adoption Below 4.5%

Microsoft 365 has roughly 450 million business users, but the adoption rate for the paid Microsoft 365 Copilot remains **below 4.5%**. Its aggressive pricing has backfired, and Copilot keeps losing ground alone while standalone ChatGPT, Anthropic's Claude, and Google's Gemini all expand their business adoption.

Companies less tied to the existing Microsoft ecosystem appear increasingly inclined to evaluate cheaper and more accurate alternatives. Whether Microsoft revisits Copilot's pricing structure and feature differentiation could shape the future landscape of the enterprise AI market.

Google Search's AI Mode Begins Integrating with Third-Party Apps

Google has started rolling out in the US a feature in Search's AI Mode that connects directly with third-party apps including **Canva, YouTube Music, and Instacart**. Users can complete tasks such as creating designs, adding items to a cart, or playing music without ever leaving search results. Combined with "Personal Intelligence," the feature is said to deliver personalized responses and actions tailored to the user's context.

The move signals a gradual expansion of search from "information retrieval" to "task execution," and stands as a symbolic example of the boundary between AI assistants and search blurring even further.

---