Google just gave Gemini access to your entire digital life, and most people are going to say yes without thinking about it.

At I/O 2026, the company rolled out Personal Intelligence, a feature that lets Gemini read your Gmail, Photos, and other Google services to give answers shaped around you. It's expanding to nearly 200 countries and reaches the roughly 900 million people who use Gemini each month.

In a demo, Google Labs head Josh Woodward asked Gemini what tires to buy for his car. Instead of generic specs, it pulled the part number from a photo buried in his Google Photos and suggested options based on how he actually drives. That kind of usefulness depends on Gemini knowing a lot about you — which Google is pretty open about wanting. "We think there's a huge opportunity for our AI to know you better," Robby Stein, VP of Product for Google Search, said on the Limitless podcast.

Here's where it gets messy. The data Google already has on you was handed over for specific reasons — Gmail for email, Photos for pictures, Maps for directions. Personal Intelligence pools all of that into a single assistant that moves information across products in ways the original sign-up never contemplated. What was once siloed is now one big blob, making it harder to draw lines around what's appropriate.

Google does offer controls, but the fine print matters:

• Connected apps are off by default, and you can disconnect them anytime
• Google says it isn't training models directly on the contents of your Gmail or Photos
• But even with chat history turned off, Google holds onto your Gemini conversations for 72 hours — and the default retention is 18 months
• The opt-in is framed as the benefit; the opt-out is tucked somewhere inconvenient — a pattern Dr. Harry Brignull, who coined the term "dark pattern", has been flagging for over a decade

The interesting question isn't whether Personal Intelligence is useful. The demos prove it is, and most people will trade some privacy for an assistant that actually remembers their kid hates museums. The question is whether 900 million people are going to think for more than a second before tapping the consent button. Probably not, because Google isn't really asking for new data here. It's asking permission to finally put a decade of data you already handed over to a use you didn't sign up for. That might be the deal of the century or the moment we stop pretending consent in tech means much of anything, and it's likely going to be both at the same time.

Microsoft finally has a personal AI assistant it's willing to call one.

On Tuesday, the company launched Scout, an always-on agent that sits in the background of your workday, watches what you're doing across apps, and starts handling things without being asked. It's built on top of OpenClaw, the open-source agent project that became the breakout AI story of 2026 — but wrapped in something the average enterprise user can actually trust.

Omar Shahine, who runs Scout, called it "the first real personal assistant we've offered customers" and said users should expect it to behave like one. The pitch is closer to having an executive assistant than running a prompt. And the honest part of the pitch is what makes Scout interesting — Shahine openly described how Microsoft treats the OpenClaw layer under the hood:

• Sandboxed by default: Scout's OpenClaw engine runs in an isolated environment on your machine, not with direct access to the rest of your system.
• Treated as untrusted: It doesn't get the keys to your Microsoft 365 data. It has to ask through controlled APIs.
• Memory that compounds: A system called Work IQ keeps building context on how you work, what you care about, and what's likely coming next.

That last one is where the cracks show. 404 Media reported on internal Microsoft documents where employees described the goal of making users "addicted" to Scout. The framing carries extra weight given that Microsoft AI CEO Mustafa Suleyman has been openly saying his goal is for Microsoft to become one of the four AI labs that matter — and Scout is the consumer-facing edge of that push.

Microsoft is doing the thing where it tells you what it's building and tells you what it actually wants from you in the same week, and somehow both can be true. Scout looks like a real step forward in personal agents, with a more grown-up security posture than most of what's shipped this year. But "always-on" and "designed to be sticky" are the same sentence in two different fonts, and the company that owns your inbox, your calendar, and now an agent that learns how you work doesn't need to be addictive to be hard to leave. The interesting question isn't whether Scout works. It's whether anyone notices when they stop being able to picture work without it.

Apple's biggest developer event kicks off Monday, and for the first time in a long time, the pressure isn't about hardware — it's about AI credibility. WWDC starts June 8, and Apple confirmed the keynote will lean heavily on software and AI. That's where the company has been bleeding goodwill for over a year. Apple Intelligence launched flat, the personalized Siri overhaul promised at WWDC 2024 slipped, and Cook told CNBC that a more personalized Siri was "on track to launch" this year. Monday is where he has to actually show it.

What makes this year different is Apple has stopped pretending it can do this alone. In January, Apple and Google confirmed that Apple Foundation Models — the brains behind Siri and most of Apple Intelligence — will run on a custom version of Google's Gemini. That's a notable concession from the most vertically integrated company in tech. As Wedbush's Dan Ives put it on CNBC: "It's certainly a win if you think about the pain that they've had in their AI strategy up to this point." The win is that Siri might finally work. The trade-off is that the thing making it work isn't theirs.

Here's what the keynote is expected to cover, based on roundups from MacRumors and CNET:

• The Siri overhaul, finally: A more conversational assistant that can pull personal context from your apps and take actions on your behalf — the feature Cook promised two years ago.
• On-device AI as the pitch: Expect Apple to lean hard on the idea that your data stays on your phone, an obvious counterweight to Gemini running in the background.
• iOS 27 and a redesign: The bigger annual software refresh, where the new Siri will live alongside whatever else Apple wants developers to build against.

Sitting under all of this is a leadership transition. In April, Apple announced that Cook will step aside and hardware engineering chief John Ternus will become CEO on September 1. That makes WWDC essentially Cook's last big keynote and Ternus's first real introduction to the audience that will judge him.

The story Apple wants to tell on Monday is that the delays were worth it, that the Gemini deal is a strategic choice rather than a concession, and that Siri is finally about to be the assistant it was always supposed to be. The story the keynote will actually tell is whether that's true. Apple has 2.2 billion active devices and a customer base that has been remarkably patient about a year of unmet promises, but patience isn't infinite, and the company is handing the keys to a new CEO in September with the AI bet still unproven. If Ternus inherits a Siri that works, the Gemini partnership looks like a smart trade. If he inherits another set of "coming later this year" slides, it's going to be a long fall.