Google May Start Using Your Emails and Photos to Target Ads
Today’s AI news cuts to something a lot of people have been quietly wondering about: when you use AI tools built into services you already rely on, what happens to your data? One answer arrived this week — and it’s not entirely comfortable.
Google Reportedly Updated Its Privacy Policy to Let AI Features Feed Ad Targeting
For years, Google drew a clear line: the contents of your Gmail and Google Photos were off-limits for advertising. According to The Verge, that line may have just moved.
Google reportedly updated its privacy policy to allow personal data from Gmail, Google Photos, and similar services to be used for targeted advertising — but only when users access AI features built into those products. So if you use Gemini to summarize an email or let Google’s AI organize your photo library, that interaction may now inform the ads you see.
Think of it like a deal you didn’t explicitly sign. You get a useful AI assistant inside a free product. In exchange, the company learns more about you from the very data it’s helping you manage. It’s a trade-off that’s always existed in some form with free software, but this update expands it into more personal territory than before.
For everyday users, this is worth paying attention to. Your inbox contains medical questions, financial conversations, and messages from family members. Your photo library holds years of personal moments. The idea that those could now influence ad targeting — even indirectly through AI feature usage — represents a meaningful shift in how Google treats that information.
Why this matters: People often assume that because a service is labeled “private,” it stays that way. This update is a reminder to check what you’re agreeing to before using AI features inside any platform you trust with sensitive data.
“Google changed privacy policy to use personal emails and photos for targeted ads.”
Anthropic’s Developer Tools Just Got Better at Organizing Work and Tracking Costs
Anthropic — the company behind the Claude family of AI models — released two updates to its Python SDK this week. An SDK, or software development kit, is essentially a toolbox that developers use to build apps and services on top of an AI model. These updates don’t change Claude itself, but they change how developers work with it.
Version 2.2.0, detailed on GitHub, added two things developers have wanted for a while: message threads and usage reporting. Message threads let developers group related conversations together — imagine building a customer service tool where each support ticket stays neatly organized rather than everything mixing into one stream. The usage reporting feature lets businesses see exactly how much of the API they’re consuming and where costs are accumulating, broken down by category.
Version 2.3.0 followed quickly, adding tag management for Claude Enterprise organizations and fixing a technical issue with Google Cloud authentication. Tags work like labels — an organization could tag conversations by department, project, or client, making it much easier to audit how Claude is being used across a large team.
For people who use products built on Claude — writing assistants, research tools, customer support bots — these updates make the underlying service more reliable and easier for developers to maintain. Better tooling means fewer bugs, cleaner experiences, and more predictable pricing for the businesses building on top of Claude.
Why this matters: Anthropic is clearly investing in making Claude easier to deploy at scale. Two SDK releases in close succession signals that enterprise customers are pushing for more control and visibility — and Anthropic is listening.
“Message threads for organizing conversations and usage reporting by category.”
Also Happening in AI
A handful of developer tools shipped updates this week. Ollama released v0.40.10 with support for NVIDIA’s Nemotron 3.5 Bolt, a model designed to run efficiently on local hardware — meaning your own computer rather than a remote server. LiteLLM, a library that helps developers connect to many different AI models through a single interface, pushed two updates: v1.106.3 with a security change to its digital signing key, and v1.105.4 with additional fixes. LangChain, a popular framework for building AI-powered applications, released core version 1.7.0, which now streams tool-call arguments as individual pieces rather than waiting for a full response — making AI apps feel noticeably faster to end users. OpenAI also updated its Python library to v3.37.0, adding support for structured instruction formatting that gives developers more precise control over how they prompt models.
What to Watch
The Google privacy story is the one to keep following. The key question isn’t just what the updated policy says — it’s how Google implements it and whether regulators in the EU, where privacy rules are stricter, push back. Watch also for whether other AI-integrated platforms quietly follow with similar updates. As AI features become standard inside apps people use every day, the data trade-offs underneath them are becoming harder to ignore.