The UK Can’t Have the New AI Agents Yet. What to Do With the Time.

Since May, four of the biggest names in AI have launched personal agents, three of them in the last seven weeks. Google’s Gemini Spark, Meta’s Muse, OpenAI’s dots and, this week, Anthropic’s Claude Duet.

They read your email, keep your diary and get on with tasks while your laptop’s shut. Duet sends you a briefing in the morning. dots runs on its own computer in the cloud.

And if you’re reading this in the UK, you can’t have any of them on a personal account.

Good news or bad? The delay buys you time. It also decides how these agents arrive: on a company seat with admin controls, or on someone’s personal account over a VPN. Right now the second route is the easier one.

What You Can and Can’t Get

One exception. OpenAI shut dots to UK Pro users, the personal plan. It’s also on ChatGPT Business Premium seats, which OpenAI says include dots in every supported region, including the UK. Those seats have a list price of $125 a user a month on monthly billing, or $100 a month billed annually, with a two-seat minimum.

So for dots there is a business route into the UK. For Muse, Spark and Duet there isn’t one.

Why Has the UK Been Left Out?

Nobody’s said.

Meta’s launch announcement doesn’t mention Europe or the UK at all. Google gave no reason for Spark. Anthropic said Duet starts with US subscribers and left it there.

Sam Altman got closest. NBC News reported him putting it down to the region’s “strict regulations”, and saying launches there “take longer, and sometimes unpredictably longer”.

Here’s the odd bit. The UK doesn’t have an AI Act. The King’s Speech in May 2026 contained no AI bill. So “EU rules” can’t be the whole story.

Apple is the test case. Apple blamed the EU’s Digital Markets Act for keeping Siri AI off iPhones in the EU, and launched it in the UK on day one in September, because that law doesn’t apply here. The companies that named no law are the ones that left the UK out.

My read is that the AI companies treat the UK as part of a European bloc because of UK GDPR and an active regulator. On 8 October the ICO said ten AI developers, including Anthropic, Google, Meta and OpenAI, had made or committed to data protection changes. It also opened a call for evidence on AI agents that closes on 20 November.

Whatever the reason, there’s no law that forces a UK date. If your plan is to wait until it launches here, you’re waiting on something nobody has promised.

How Long Has the UK Waited Before?

We’ve been here before. How long did the UK wait last time?

For a while, the gap was closing. Sora took 81 days to reach the UK, and some of OpenAI’s later features reached Europe within a week.

Meta AI, which wanted the same kind of access to your accounts, took 174 days to reach the UK and 336 to reach the EU. Gemini Spark is already at 158 days with no date.

This year’s agents want deeper access than anything before them. Muse lets you choose read-only or read-and-send for your email. Duet checks Gmail and Slack every few hours. Nothing in that history says this one will be quick.

The ICO itself said in January that personal shopping agents would arrive “within the next five years”. They arrived within nine months.

Will Your Team Wait?

Some won’t. And UK people have form.

When age checks came in under the Online Safety Act on 25 July 2025, Proton VPN saw sign-ups jump 1,400% in the first hours and stay around 1,800% above normal for several days. By 28 July it was the UK’s most downloaded free iPhone app, ahead of ChatGPT.

It’s happened with AI too. Before ChatGPT’s Advanced Voice reached Europe in October 2024, The Decoder reported that access from those countries “was only possible via VPN”.

For this year’s agents, step-by-step guides on using Muse through a VPN were already online by 12 October. Nobody has measured how many UK people are doing it yet, so I won’t pretend to know. The precedent is strong.

And it breaks the rules. OpenAI’s help pages say using ChatGPT from an unsupported country “may result in your account being blocked or suspended”.

Microsoft found in October 2025 that 71% of UK employees have already used unapproved consumer AI at work. More on that in our shadow AI post.

Why an Agent on a Personal Account Is a Different Problem

Connect Muse or Duet to a work inbox and it reads the whole inbox, customer names included, all day, on the vendor’s computers. The security side is in our shadow AI post. The legal side is this.

This is my own reading of the rules, so treat it as that. If someone connects a personal agent to a work mailbox, the business is still responsible for the personal data in that inbox. But it has no contract with the agent’s provider, no data protection assessment and no say over how long data is kept or what it’s used for. Axios reported Muse’s default setting lets Meta use your conversations to train its models, unless you opt out.

The ICO’s current position, as summarised by Freshfields on 22 October, is that neither the provider’s terms nor a user’s instruction to an agent gives you a lawful basis for everything the agent does. It also expects human approval for high-risk actions and access kept to the minimum needed.

None of that is set up when someone connects their own agent from home on a VPN.

What to Do Before They Land

They can arrive on a business seat, with admin controls, a contract and someone accountable. Or they arrive on a personal account over a VPN, and the first you hear of it is when something goes wrong.

Nobody knows how long the UK has. The history says months. That’s long enough to get the rules and the training in place, if you start now.

So when the first person on your team connects an agent to their work email, will you know about it?

Sources & References

Frequently Asked Questions

Roy Murphy is co-founder of G3NR8. This post draws on launch announcements, regulator statements and reporting from 2023 to October 2026. Published 24 October 2026.