Big Tech in June 2026: the light and the dark

Good evening. The Owl here.

For GAFAM, June 2026 was the month AI stopped being the star of the launch event and moved into the front doors of work, search, smartphones, advertising and the cloud.

At WWDC26, Apple put Siri AI front and center, and at Build 2026, Microsoft announced AI agents for work. AWS expanded the tools that make it easier for companies to run AI agents in production, and Meta, too, is building AI deep into Facebook search and its services for businesses.

At the same time, June was a month that put AI’s permissions and responsibilities under scrutiny. The UK’s CMA imposed publisher-protection requirements on the AI features in Google Search, and Apple delayed Siri AI in the EU. Reports of unauthorized access through Meta’s AI support, and a facial recognition lawsuit against Amazon’s Ring, also show that the more convenient AI becomes, the more weight falls on how its boundaries are designed.

---

The light side

Apple puts Siri AI front and center at WWDC26

At WWDC26 on June 8, Apple announced the next generation of Apple Intelligence and the new “Siri AI.” Siri AI is described as an assistant that is more conversational, understands your personal context, and helps you across mail, messages, photos, actions inside apps and more.[1]

This announcement matters a great deal for Apple. Apple Intelligence has been rolling out in stages since 2024, but what users were hoping for most was, unsurprisingly, a rebuilt Siri. This time, rather than setting AI apart as a separate app, Apple reaffirmed its direction of building it deep into the iPhone, iPad, Mac, Apple Watch and Vision Pro.

Apple’s strength isn’t the name of a cutting-edge model. It’s the devices people touch every day and the access its operating systems have. Not many companies can place AI right next to your photos, calendar, mail, browser and Shortcuts. If Siri AI really works reliably, Apple can compete not as “the smartest AI” but as “the AI closest to your daily life.”

That said, this is also an announcement about winning back a promise. Siri has long been called a disappointment. What matters more than the applause at WWDC is whether, from the fall onward, people actually use it and feel, “Yes, I can trust this with my everyday life.”

Microsoft puts AI agents at the front door of work

At Build 2026, Microsoft announced “Microsoft Scout,” a personal agent for work. Scout is built on OpenClaw and WorkIQ and is described as understanding everyday work tools such as Teams and Outlook, and handling meeting prep, scheduling and routine tasks before you even ask.[2]

Microsoft’s aim is easy to read. It wants to put AI into the flow of work itself, not treat it as a writing tool. Precisely because meetings, email, calendars, internal documents and task management all come together in Microsoft 365, the company is well placed to own “the context of work.”

If AI can read the materials before a meeting, spot scheduling conflicts, draft replies and move the next task forward, the way office workers spend their time will change dramatically. With less busywork, people can more easily spend their time on judgment and creative work.

At the same time, permission management is the crux of AI agents for work. Reading email, moving appointments, touching files, contacting someone. Behind the convenience, there will always be the question of how to prevent mistaken actions and information leaks.

AWS pushes AI agents into the enterprise rollout stage

At AWS Summit New York 2026 in June, AWS announced new features that make AI agents easier to use at work. The announcements, including Amazon Bedrock Managed Knowledge Base, AWS Context and AWS Continuum, spanned enterprise data, development and security.[3]

What AWS is emphasizing isn’t the models themselves so much as “the foundation for running AI inside a company.” Knowledge bases for RAG (retrieval-augmented generation), context management for handling internal data, agents that deal with code vulnerabilities. None of these are flashy chat screens, but they are all parts a company needs when it brings AI into real work.

The problem that often comes up in AI adoption isn’t that the model is unusable. It’s that the model can’t be connected to the company’s data, permissions and workflows. AWS is stepping into exactly that gap, trying to make AI agents part of business applications on top of its cloud platform.

This move is also a new form of cloud competition. The more AI agents companies build on AWS, the more their data, authentication, monitoring, billing and development environments gravitate toward AWS. AI is also becoming a front door for more cloud usage.

Meta extends AI into Facebook and customer service for businesses

In June, Meta announced it would bring AI Mode to Facebook. It’s a search tab that returns answers from Meta AI, not just links, drawing on the culture, opinions and recommendations found in public posts. Meta also announced AI features that help people create photos and videos.[4]

Meta also announced “Meta Business Agent” for businesses. It’s described as an AI that lets businesses automate customer service and can also connect to their existing infrastructure.[5]

What’s very Meta about this is the effort to tie AI to everything: search, ads, customer service and creative work. Facebook and Instagram bring together users’ posts, store information, advertisers, creators and messaging. Once AI is woven in, search, customer service and advertising can all more easily be completed inside the same platform.

This looks like a promising service, especially for small businesses, because even when they are short-staffed, they can use AI to cover inquiries, product explanations and the path to booking. On the other hand, it also means handing even the front door of customer service to the platform. Here, too, convenience and dependence are two sides of the same coin.

The dark side

UK CMA requires publisher protections in Google’s AI search

On June 3, the UK’s Competition and Markets Authority (CMA) introduced new conduct requirements for Google’s search services. They give publishers more control over whether their content can be used in the AI-generated features in Google Search, and require clear links and attribution when publisher content is used in AI-generated results.[6]

This is a move into a problem at the very root of AI search. In traditional search, Google created value by sending users to outside sites. With AI search, though, the answer tends to be complete right on Google’s own page.

For users, that’s convenient. But from the point of view of the people who wrote the articles, their content is used as raw material for AI while visits may fall. If advertising revenue and the relationship with readers weaken, the people producing high-quality information will be worn down.

The CMA’s decision doesn’t stop AI search. Rather, it’s an intervention that says: if AI search is going to spread, the publishers and site owners who create its raw material should keep some bargaining power.

Apple delays Siri AI in the EU

Even as it unveiled Siri AI at WWDC26, Apple announced that it would not be able to offer Siri AI in the EU when iOS 27 and iPadOS 27 are released. It cited compliance with the Digital Markets Act (DMA) as the reason.[7]

Apple explained that because EU regulators did not accept its proposals, including how it would accommodate other virtual assistants, it cannot offer Siri AI safely. In short, the deeper an AI assistant reaches into the operating system, the more easily competition policy and security design collide.

What’s at stake here isn’t just Apple’s own interests. If an AI assistant touches users’ mail, calendars, screens and in-app actions, should Apple alone hold that access? Or should the same front door be opened to competing assistants?

Open it, and competition increases. Keep it closed, and it’s easier to preserve safety and an integrated experience. The delay of Siri AI in the EU shows that in the AI era, the operating system has become not just a matter of “useful features” but a central theme of regulation.

Meta’s AI support problem shows the danger of handing over permissions

In June, Reuters reported that attackers had tricked Meta’s AI support chatbot and gained access to high-profile Instagram accounts. The report pointed to the risks of using AI to automate sensitive functions such as account recovery.[8]

What’s frightening about this problem is that the AI didn’t simply give a wrong answer. Account recovery is the area where identity verification, impersonation countermeasures and fraud detection matter most. If you put AI there, and connect it to processes that carry real permissions, then designing the lines it must never cross becomes more important than its ability to hold a natural conversation.

An AI agent that only reads is still relatively safe. Writing, sending, changing, recovering, approving, paying: once it starts touching operations like these, the impact of mistakes and abuse grows all at once.

In June, the GAFAM companies showed us a future where we hand our work to AI. That future also has pitfalls like the one in this Meta story.

Amazon’s Ring sued over facial recognition

Amazon-owned Ring was also hit with a class action lawsuit in June over its facial recognition feature, “Familiar Faces.” According to Ars Technica, the lawsuit claims that Ring cameras scanned the faces of visitors and passersby and handled facial recognition data without those people’s consent.[9]

Home camera owners may turn the feature on for security or convenience. But the camera doesn’t capture only its owner. Delivery workers, neighbors, passersby, friends: their faces may be recognized without them ever knowing.

Once AI starts looking at the physical world, “the user consented, so it’s fine” is no longer enough. Unlike data inside a phone or PC, footage of a front porch or a street includes people who aren’t users of the service at all.

The Ring lawsuit shows that the more AI cameras and smart homes spread, the more the boundaries of privacy seep out beyond the home.

---

Looking back at GAFAM in June 2026, AI has come down to earth in some very practical places. Apple is trying to remake Siri with AI, Microsoft work, AWS enterprise systems, Meta search and customer service, and Google the front door of search.

But the more AI becomes something we can “hand things over to,” the heavier the question of permissions becomes. What do we let it read? What do we let it operate? Which data do we let it use? Whose consent is needed?

June’s light is that AI has begun, little by little, to take some of the hassle out of our lives and work. The dark is that this convenience is starting to swallow up the front doors to information, permissions in the workplace, personal data, and even footage of the physical world.

🦉 What I’m watching, more than AI getting smarter, is the moment we hand AI the keys. The more convenient doors there are, the more I want to keep in mind who holds those keys.

References

[1] Apple: WWDC26: Apple unveils next generation of Apple Intelligence, Siri AI, powerful parental controls, and an expansive set of software improvements

[2] Microsoft Official Blog: Microsoft Build 2026: Be yourself at work

[3] Amazon: AWS Summit New York 2026: New ways to make AI agents more effective at work

[4] Meta: New AI Tools to Help You Make Things Happen on Facebook

[5] Meta: Be There for Every Customer With Meta Business Agent

[6] UK Government: CMA secures fairer deal for publishers and improves Google search services in UK

[7] Apple: Due to DMA, Siri AI delayed in EU for iOS 27 and iPadOS 27

[8] Reuters: High-profile Instagram AI chatbot breach spotlights security risks of automation

[9] Ars Technica: Amazon-owned Ring should pay Americans for scanning their faces, lawsuit says