The US tech that is not reaching Europe

What people in Italy are missing

On 13 October Apple unveiled HomeView, a display for the home that plays video, makes video calls and controls lights and thermostats by voice. The announcement has one line that concerns us: HomeView is not for sale in the European Union, because Siri AI is not available on iOS here.¹ Siri AI is the new version of Apple's assistant, presented on 8 June. In the EU it works on Macs and on the Vision Pro headset, but not on iPhones and iPads, and so not on the Apple Watch either, which needs a paired iPhone.²

It is not the only case. In January Google launched Personal Intelligence, which lets Gemini look for answers in your Gmail, Google Photos, and YouTube and search history. On 20 October it brought the feature to cars, through Android Auto, everywhere in the world except the European Economic Area, Switzerland, the United Kingdom and Nigeria.³ The same exclusions apply to Gemini Spark, the agent that carries out tasks in the background, and to Gemini's scheduled actions. Chrome's feature that uses the browser on your behalf, auto browse, exists in the United States and, since 7 October, in India, but in no European country.⁴ On 8 September Meta launched Muse, a personal agent, in North America, and on 23 October it brought it to Japan. For the small device built to talk to it, Muse Charm, shown on 23 September, there is still no price and no date for any country.⁵

Then there is health. When OpenAI announced ChatGPT's health section in January, it left out the European Economic Area, Switzerland and the UK, and the link to medical records only works with US healthcare providers.⁶ Since July the feature has been open to all American adults. In the United States Amazon offers Health AI, an assistant that explains test results, manages prescriptions and hands the conversation to One Medical doctors. As of 27 October neither company had announced plans for Europe.⁷

What is not reaching Italy, as of 27 October 2026

Service	Company	Announced	Available in Italy	Stated reason
Siri AI on iPhone and iPad	Apple	June 2026	No, only on Mac and Vision Pro	The Digital Markets Act, according to Apple
HomeView	Apple	October 2026	No	No Siri AI on European iPhones
Gemini Personal Intelligence	Google	January 2026	No, as in the UK and Switzerland	None
Gemini Spark and scheduled actions	Google	May 2026	No, as in the UK and Switzerland	None
Auto browse in Chrome	Google	January 2026	No	None: only in the US and India
Muse	Meta	September 2026	No	None, no date
Health in ChatGPT	OpenAI	January 2026	No, as in the UK and Switzerland	No official reason
Health AI	Amazon	January 2026	No	Relies on One Medical, which operates in the US

The rules matter, but in different ways

There are at least three stories in this list, and it helps to keep them apart.

The first is about the Digital Markets Act, the DMA, the EU law that since 2024 has required the largest platforms not to favour their own services and to open their systems to competitors. For Apple the law covers iOS, iPadOS, the App Store and Safari, but not macOS.⁸ That is why Siri AI runs on the Mac and not on the iPhone. Apple claims the Commission wants it to give any rival assistant almost unlimited access to the phone, from messages to purchases. It says it proposed an intermediary, the Trusted System Agent, to be phased in over 18 months, and that the proposal was rejected.² On 9 June the Commission's spokesperson Thomas Regnier replied that the decision not to roll out Siri AI "is Apple's and Apple's only", that nothing in the DMA stops Apple from launching new products, and that instead of looking for a compliant solution Apple asked for an exemption.⁹ In its quarterly filing with the SEC, the US securities regulator, Apple writes that interoperability requirements have already led it, and may lead it again, not to launch products "such as Siri AI" in some jurisdictions.²

Google is on the other side of the same fight. On 16 July the Commission ordered it to give rival assistants the same access to Android features that Gemini enjoys, starting with voice activation, like "Hey Google". Users should see the changes from July 2027.¹⁰ On 28 September Google appealed to the EU General Court, arguing that the measures weaken security and privacy.¹¹

The second story is about personal data. Google's and OpenAI's exclusions also cover the UK and Switzerland, which do not apply the DMA but have privacy laws similar to the EU's GDPR. Personal Intelligence, Spark and ChatGPT's health section work on email, photos and medical records. The companies do not explain the exclusion, but the line they drew matches the reach of European-style privacy laws. Meta was more open: in July 2024 it announced that its Llama model able to read images would not come to Europe because of "the unpredictable nature" of European rules.¹² In those months the Irish Data Protection Commission, which supervises Meta for the whole EU, had raised objections, and Meta had paused training its AI on public Facebook and Instagram posts. After an opinion from the European data protection authorities, Meta added a simpler objection form, data filters and other safeguards, and since 27 May 2025, the date it gave the regulator, it has been training its models on public content from European adults too.¹²

The third story has little to do with European rules. Amazon's assistant exists to steer people towards One Medical, a network of American clinics: in Europe it would have no doctors to hand them to.⁷

What the figures say about delays

The research centre GovAI compared 375 model releases by Meta, Google, OpenAI and Anthropic between June 2018 and May 2026. Of these, 89% reached Europe at the same time as the United States, while 11% arrived late or never, against 7% in the UK, and the median delay in Europe was 53 days. Of the 68 cases, the authors attribute 56 mainly to regulatory barriers, largely data protection and especially training models on personal data. They find no strong evidence that the AI Act caused delays, partly because its rules on general-purpose models can only be enforced with fines from August 2026, after their data ends.¹³

The most useful figure is the trend. The share of releases that reached Europe late or never fell from 28% in 2023 to 8% in 2025, and in the first five months of 2026 no public model was held back. The only company going the other way is Meta, which went from 7% to 88%.¹³

AI models that reached Europe late or not at all

Four companies, 202328%

Four companies, 20258%

Meta, 20237%

Meta, 202588%

The authors note that the UK, with very similar privacy laws, had fewer problems, and put this down to stricter enforcement and slower guidance from European authorities. Recent history also shows that delays usually end, though not every time: at the end of June 2026 Google's AI Overviews had still not reached France.¹³

Delays that ended, and fights still open

1. Jul 2023Bard, Google's chatbot, reaches Europe after questions from the Irish regulator
2. Dec 2023Threads reaches Europe, five months after launch
3. Jun 2024Apple postpones Apple Intelligence in Europe, citing the DMA
4. Mar 2025Meta AI reaches 41 European countries; AI Overviews reach Italy; Apple Intelligence reaches European iPhones, Italian included
5. May 2025Meta starts training its AI on European public posts, with new safeguards
6. Oct 2025AI Mode, Google's conversational search, reaches Italy
7. Jun 2026Siri AI is kept off European iPhones
8. Jul 2026The Commission orders Google to open Android to rival assistants
9. Oct 2026HomeView is not put on sale in the EU

The Italian case

For people in Italy the list of missing features matters less than another figure. In 2025, 19.9% of Italians aged 16 to 74 used a generative AI tool, against an EU average of 32.7%. Only Romania did worse. Among companies with at least ten employees, 16.4% use AI, twice as many as a year earlier, with a huge gap between large companies and small and medium ones.¹⁴

For an Italian company starting out, the tools on offer are more than enough, and the main obstacle, as companies themselves say, is skills. But that 47.3% shows that uncertainty about the rules slows down more than just Apple or Google, and Italy has first-hand evidence of it. On 30 March 2023 the Garante, Italy's data protection authority, blocked ChatGPT, which came back at the end of April. In November 2024 it fined OpenAI 15 million euros, and on 18 March 2026 the Court of Rome upheld OpenAI's challenge to that fine.¹⁵ For a company deciding whether to invest, three years of decisions that get overturned weigh heavily.

For people at work the gap is smaller than it looks. Many of the missing features are built for personal accounts: Personal Intelligence, for instance, is not available for Google Workspace business accounts even in the US.³ And developers often get access to models from day one: GovAI recalls that Claude 3 Opus reached the European app 71 days late, while access through the API came with no wait.¹³ Anyone at work tempted to reach a service not available in Italy through a VPN should stop: it would put company data into a service its provider chose not to offer on the European market and that nobody has assessed against European rules. I wrote about the rules that already apply in Italy, and who enforces them from 23 October, in Who will police AI in Italy.

What I think

I do not think you can side wholesale with the companies or with Brussels. You have to look at one case at a time and ask what the rule would protect if it worked.

On health data and on training with personal data, European caution protects people. An assistant that reads medical records, email and photos can know more about us than a family doctor, and that data, once it is out, does not come back: in another piece I described how what people confide to a chatbot can end up for sale. If a service like that arrives a few months later because a regulator asked questions and the company had to answer them, I think that is an acceptable price. The Meta story shows it: just over a year of discussion produced a simple way to object, and then the service started.

On Siri AI, though, I think the absence is also a bargaining chip. In June 2024 Apple said Apple Intelligence would not come to Europe that year because of the DMA, and in March 2025 it arrived, in Italian, without the law having changed.¹⁶ In September 2025 Apple asked for the DMA to be repealed.¹⁷ Siri AI works on Macs sold in Europe: the technology is not banned, what Apple will not accept is giving rivals the same access on the iPhone. Giving up on selling a whole product, HomeView, over a software dispute is a way to pass the cost on to customers and push them to put pressure on Brussels. Apple is entitled to do it, but the decision is still its own. That one company can decide on its own what reaches a continent is exactly the kind of dependence I wrote about in Who holds the keys?.

None of this lets the Commission off. The Draghi report counted around 100 EU laws focused on tech and more than 270 regulators dealing with them across member states, and noted that only larger companies, often not European, can bear those costs, while young tech companies may choose not to operate in the EU at all.¹⁸ The rules weigh more on people building things in Europe than on Apple. Dirk Auer, of the International Center for Law & Economics, points out that a law designed to increase competition here does the opposite, because it takes a competitor off the market.¹⁹ I do not fully agree, but the point is a serious one.

I would ask the Commission to decide faster and with clear technical criteria. With Google it set out the features, the security safeguards and a year to comply: the same should be done with Apple, with public milestones that can be checked, instead of a war of press releases. On the data part of the Digital Omnibus, proposed in November 2025, I would be in no hurry at all. It would write into law that legitimate interest is enough to train AI on personal data, with safeguards and a right to object: it means each of us has to say no, instead of companies having to ask for a yes. It is still being discussed in European Parliament committees, with no Council position, and that is where it should be made stricter, not rushed through.²⁰ The companies, for their part, should say why they leave a service out. Google keeps Personal Intelligence and Spark out of Europe without a line of explanation: one sentence would help users and authorities understand whether the problem is a rule, a legal risk or a commercial choice.

There are two things to watch next: the EU General Court's ruling on Google's appeal, and July 2027, when rival assistants must be able to work on Android just as Gemini does.¹⁰