Linked List: June 15, 2026

Monday, 15 June 2026

The Washington Post on the EU’s DMA Folly 

The Washington Post editorial board yesterday (News+ link), “Why Europe Won’t Have the New Siri”:

Brussels insists the decision is “Apple’s and Apple’s only” and that nothing in its flagship Digital Markets Act forbids the launch. That’s technically true and wholly beside the point.

The law requires that the moment Siri AI ships in Europe, any rival AI agent must get the same sweeping access to a user’s messages, files and chat history. Apple proposed putting in a software security layer to make that safe and offered a phased rollout to build it. According to Apple, the European Commission rejected the proposal.

The DMA was supposed to open markets. But its legal logic was born in the era of browsers, app stores and messaging apps. These components can be swapped like batteries.

The DMA effectively demands everything to be swappable/interchangeable. So while the European Commission is correct that the DMA does not forbid Apple from launching a version of Siri AI, it clearly forbids Apple from launching the version of Siri AI they actually built.

Behind all this lies the dream that Europe could be a “regulatory superpower.” It wanted to create a market too big to skip that would, by virtue of its heft, end up exporting its rules to the rest of the world. That hasn’t worked out.

When adapting a product for Europe costs more than European market access is worth, companies no longer comply. They simply leave out the feature.

That’s the folly of the DMA, or at least the maximal interpretation of the DMA that the European Commission is pursuing. It only makes sense under the assumption that the EU is too big a market to ignore, and the EU’s market might is such that systems will be designed to meet their compliance standards, regardless of whether the makers of these systems support the regulations or not. (And in the case of Apple with iOS and Google with Android, the two companies are in lockstep in their opposition to the EU’s regulations on system-level AI interoperability.)

First, the EU is big but it isn’t that big. The best estimate I’ve seen is that the EU accounts for about 7% of Apple’s worldwide revenue. Plus, because of the DMA, the cost of doing business in the EU is now significantly higher for Apple and Google, because they need to engineer DMA-compliant versions of various features and systems. Unless, that is, they stop bringing (a long and ever-growing list of) new features to the EU.

Which brings me to my second point. What exactly is the motivation for Apple and Google to engineer entirely separate systems for the EU to bring new features into compliance with the Commission’s broad interpretation of the DMA? Because if Apple doesn’t engineer a DMA-compliant version of Siri AI, iOS users in the EU will ... switch to Android, whose system-level AI was deemed noncompliant by the Commission a few months ago?

This doesn’t directly hurt Apple. It doesn’t force Apple to design, engineer, and ship a compliant EU-exclusive version of Siri AI that supports plug-and-play LLM back ends. It only hurts iPhone users who live in the EU, who are stuck with the old dumb version of Siri for the foreseeable future. The European Commission is either stupid or insane.

The European Commission Ruled Months Ago That Google’s Integration of Gemini in Android Violates the DMA 

Ryan Whitham, writing for Ars Technica back in April:

European regulators are proposing several broad changes to the way AI tools operate on Android phones. Some of this is straightforward, like allowing third-party AI tools to be invoked system-wide via hot words or button presses. This might also include allowing AI tools to view screen context when the user opens them. Context also extends to allowing alternative AI systems to access local data to generate proactive suggestions and summaries. The report actually describes something that sounds like Google’s Magic Cue, which relies on Gemini to offer suggestions based on your activity.

Google has also started experimenting with allowing AI to control certain apps. As we saw when this feature debuted on the Galaxy S26, Gemini is currently pretty bad at using apps on your behalf. The commission wants to explore allowing other AI services to autonomously control installed apps and system features on Android phones. Maybe someone else could do better?

Maybe! But also maybe it’s a bad idea for complex system architecture design to come from non-technical government bureaucrats. One of these maybes strikes me as a lot more likely than the other.

Many of the Gemini AI features in Android, including Magic Cue, rely on running local models, and Google has been slow to allow third parties the system access to make that work effectively. So the EU is also suggesting a mandate that would ensure developers have the necessary hardware access to run local models “with high levels of performance, availability and responsiveness.”

What could go wrong?

Finally, Google may be required under the DMA to create new APIs and offer technical assistance to other AI makers who want to plug into Android. The commission also specifies that these tools must be made available free of charge.

Of course, it’s not free of charge to provide technical assistance to one’s competitors. It’s actually a great expense.

Here’s the European Commission, announcing these “preliminary findings”:

The proposed measures aim to ensure that competing AI services can effectively interact with applications on users’ Android devices and execute tasks accordingly, such as sending an email using the user’s preferred email app, ordering food or sharing a photo with friends. Currently, Google largely reserves these capabilities for use by its own AI offerings on Android phones and tablets. For example, the measures would allow competing AI services to be easily activated by users, using a custom ‘wake word’, a phrase that the user can speak to activate an AI service.

The proposed measures will also enable competing providers of AI services to innovate and offer deeply integrated AI experiences to users on Android phones and tablets, along with Alphabet’s own AI services, such as Gemini. Opening up access to these capabilities will provide Android users across the EU with a wider choice of AI services.

The difference between Google and Apple on this front is that Google just blazed ahead and shipped Gemini integrated into Android in the EU, and is now facing compliance problems after shipping. (Ask forgiveness.) Apple isn’t shipping Siri AI in the EU in iOS 27, knowing that it’s going to be deemed non-compliant. (Ask permission.)

The EC presumes that these measures “will also enable competing providers of AI services to innovate and offer deeply integrated AI experiences to users on Android phones and tablets”. Again: maybe! But really all they can enforce is that “competing providers of AI services” will have the same level of system-level integration that Google’s AI services have. The easiest way for Google to achieve that is by withdrawing Gemini integration in Android from the EU, not by building APIs and privacy protection mechanisms to enable the capabilities for third-party providers that the EC is demanding.

Google is learning the lesson Apple learned the hard way with all the existing features of iOS that were deemed noncompliant with the DMA when it went into effect. The “ship it first and ask forgiveness / hope it’s deemed compliant” strategy is not a good one in the EU.

WorkOS Launches Auth.md – an Open Protocol for Agent Registration 

My thanks to WorkOS for sponsoring DF last week to promote Auth.md, their new open protocol for AI agent registration.

Sign-up forms were built for humans in browsers, so how do AI agents programmatically register with services? That’s the question Auth.md aims to answer. By exposing a single, machine-readable Markdown file at your service root, AI agents can dynamically discover your OAuth Protected Resource Metadata, parse required scopes, and authenticate seamlessly.

Markdown, baby. Who’d have thunk it?

With native support in WorkOS AuthKit, you can now implement this protocol out of the box, giving AI tools a standardized, secure way to log into your application. Read the Auth.md docs, and watch its on-stage introduction at the MCP Night: Agent Night keynote.

‘Anthropic’s Safety Superpower’ 

Ben Thompson, in his weekly free column at Stratechery:

On one hand, I actually don’t begrudge Anthropic not wanting to help its competitors; on the other hand, what should be blisteringly clear is that Anthropic does not think that anyone else other than them should even be making frontier LLMs.

What makes this policy all the more remarkable is the fact that it was enacted only two months after Anthropic had that dispute with the Department of War: the latter wanted to use Claude for any legal use, while the former wanted more stringent controls around surveillance and autonomous weapons. What this degradation represented was both the capability and willingness of Anthropic to silently alter its models to achieve its policy preferences. In other words, Anthropic willfully validated some of its critics’ worst fears in terms of being a supply chain risk.

The broader takeaway from that previous episode, however, is that Anthropic believes that they are the ones who should have final say over how Anthropic is used; given that they think only they should be developing leading edge AI, they by extension think that only they should have final say over AI generally. When you further combine this realization with the company’s pronouncements about AI’s ability to conduct all economic activity, you realize that Anthropic’s leadership effectively wants to have power over everything and everyone.

Anthropic is best seen as a religious organization. Their employees are true believers in a cause, and on a mission. Perhaps every successful company has a religious aspect at its core — like, maybe, Apple’s is design quality and user-centricism, Microsoft’s is market share with no regard for technical or design elegance, Google’s is market share with high regard for technical elegance, and Meta’s is strip-mining the world’s social graph for profit. These companies tend to attract employees who believe in the company’s core mission, and the employees who believe tend to be the ones who thrive and rise within the companies’ ranks to positions of influence.

But Anthropic feels more like a real religion, where the core tenets must be taken on faith, and the priests (Anthropic employees) have a conviction about them. A religious fervor. If Apple gets too taken away by its cultural fervor for design, they do something silly like make a $20,000 solid gold Apple Watch. So what? If Microsoft or Google get taken away by their shared fervor for market share at all costs, they face antitrust remedies. A stifled market and abusive behavior from a monopolist isn’t good, but doesn’t end the world.

A religious fervor that believes the company is building god-like “super intelligence” that will dwarf human intelligence — and that only the company’s priesthood can be trusted to define, create, control, and gate access to it — is something else entirely. I tend to think the Anthropic true believers are all wet — that LLMs, amazing though they are, are not a path toward “super intelligence”. But, they used to be clearly behind OpenAI in technical capability, then caught up, and now with Mythos/Fable, they are clearly ahead. I still think they’re wrong about where this is heading, but I don’t think we can say we know they’re wrong.