Why Sweden blocked ChatGPT and Claude, and what "no agreement" means for your data

The Morse teamPublished October 27, 20269 min read

Sweden's government blocks ChatGPT and Claude only on its own staff's work computers, not for the country. From Monday, October 26, 2026, the Government Offices (Regeringskansliet), the ministries that serve the prime minister and the cabinet, cut off the browser versions of both and kept Microsoft Copilot as their approved AI tool. The reason given is that using AI services "without an agreement" can have "consequences for our information security," in the words of digitalization director Mikael Wettercrantz, as reported by Sveriges Radio's Ekot. The government didn't drop the models either. Staff can still reach ChatGPT and Claude through an internal platform.

We read Ekot's original report, the follow-ups from SVT, Computer Sweden, and Breakit, and the official statement the Government Offices emailed to SVT, and translated the Swedish quotes ourselves. Then we read OpenAI's, Anthropic's, and Microsoft's own terms, privacy policies, and help pages on October 27, 2026, to see what a missing agreement changes. We found no statement about the block on regeringen.se, so every detail of it below comes from a named news report. OpenAI and Anthropic publish their business terms in full, and both rule out training on business customers' data by default.

What did Sweden's Government Offices block?

The browser versions of ChatGPT and Claude, on work computers. In an email to SVT, the Government Offices wrote (our translation): "The browser interface for the AI services ChatGPT and Claude will no longer be reachable from the Government Offices' work computers from October 26. Within Intric, employees still have access to the language models ChatGPT and Claude" (SVT, October 23, 2026).

Intric is a Stockholm company. On July 8, 2026, it announced that it had won the contract to deliver an AI platform to the Government Offices, one that "runs on Swedish infrastructure without dependence on foreign cloud services" (Intric). Its announcement doesn't name the models. Ekot describes the platform differently, as one for AI assistants "that the Government Offices developed themselves," and says only "some of the staff" keep access to it (Ekot, October 23, 2026). Neither report says how many people that is.

So the block closes one route to the models, the consumer website. Some English-language headlines called it a ban on ChatGPT and Claude. Ekot adds that, according to internal information it saw, the Government Offices also "raised the security" of Copilot after a review, and that the press office wouldn't give details because they concern security. The Government Offices had already restricted DeepSeek before this, Ekot says.

Why did Sweden block ChatGPT and Claude?

Because staff could use them with no contract between the Government Offices and OpenAI or Anthropic. Wettercrantz declined an interview with Ekot, but wrote on the staff intranet that using AI services "utan avtal" (without an agreement) can have "konsekvenser för vår informationssäkerhet," consequences for our information security (Ekot). Computer Sweden and Breakit carry the same quote, both crediting Ekot.

Ekot also quoted an internal audit at Trafikverket, the Swedish Transport Administration, which earlier in 2026 found widespread, risky use of about ten AI services: "We have seen several examples of problematic use of open AI services. The examples concern both the sensitivity of the information that was entered and how the AI-generated answers were handled." Some English-language reports attribute that quote to Wettercrantz. In Ekot's story it comes from Trafikverket's auditors.

What does "no agreement" mean in practice?

It means the AI company's contract is with the employee alone. A civil servant who signs up for ChatGPT or Claude accepts consumer terms as an individual. The Government Offices aren't a party, so they get none of the promises a business contract carries.

The contract decides who is in charge of the data. For people in the EEA, OpenAI's European privacy policy names OpenAI Ireland "the controller" of a consumer account's data (OpenAI), and Anthropic's privacy policy names Anthropic Ireland (Anthropic). The AI company sets the purposes, the retention, and the training rules, and the employer has no say.

A business contract turns that around. OpenAI's Data Processing Addendum, effective January 1, 2026, says OpenAI "acts as a Data Processor on the Customer's behalf" and will "only Process Customer Data for the purposes of delivering the Services to Customer" (OpenAI). The table below sets out what that and Anthropic's addendum promise the customer. Anthropic's Commercial Terms say "Anthropic may not train models on Customer Content from Services," treat that content as the customer's confidential information, and incorporate Anthropic's own DPA (Anthropic).

None of this reaches a personal account. OpenAI says it signs a DPA "for their use of ChatGPT Business, ChatGPT Enterprise, and the API" (OpenAI, updated January 8, 2026).

Can you use a personal ChatGPT account for work?

OpenAI's terms allow it, with an addendum that protects OpenAI rather than your employer. Its European Terms of Use cover "services for individuals," including "personal, non-commercial use," and end with a "Business use of the Services addendum" that applies "if you use our Services for commercial or business use" (OpenAI). The addendum caps OpenAI's liability, adds an indemnity "if you are a business or organisation," and moves governing law to California. It adds no data processing terms. The main terms still say OpenAI "can use your Content worldwide to provide, maintain, develop, and improve our Services," with training subject to an opt-out in your settings.

Whether your employer allows it is a separate question, and only your employer can answer it. Follow its rules for anything that belongs to work. This isn't legal advice. On a work laptop, your employer may also see some of what you type into a personal account, and our guide to what your employer can see on a work laptop goes through each layer.

Personal account or business contract: what the employer is promised

Every cell comes from the company's own page, read October 27, 2026. "None" means the employer has no contract with the company, so nothing is promised to it.

With a personal account, the employer has nothing to hold the AI company to: no breach notice, no word of a request from the police, and no deletion when the relationship ends. For which AI companies train on personal accounts by default, see which AI companies train on your data, and for what an admin can read, see what your employer can see on a work laptop. The personal Copilot app's rules on training, ads, and history are in is Copilot private.

Does Copilot use Claude anyway?

It can, if an administrator turns it on. Microsoft's page on Anthropic models, dated October 23, 2026, the day Ekot broke the story, says Anthropic is now a Microsoft subprocessor and that Claude models run in Microsoft Copilot, Researcher, Copilot Studio, and Copilot in Microsoft 365 apps under Microsoft's own Product Terms and DPA (Microsoft Learn). In the EU, EFTA, and the UK, the setting starts at No users, and an AI Administrator or Global Administrator can switch it on.

The same page adds two exceptions. Anthropic models "are currently excluded from the EU Data Boundary," Microsoft's commitment to store and process customer data inside the EU and EFTA. And models labeled "Anthropic models with Data Retention" fall outside Microsoft's contract: for those, "Anthropic (not Microsoft) stores most inputs and outputs for up to 30 days before deleting them."

The Government Offices haven't said whether Claude is switched on in their Copilot, and their press office wouldn't describe how they tightened it (Ekot). Between Intric and Copilot, staff may still reach Claude's models at work. The block removes the consumer website and the consumer terms that come with it.

How Morse keeps your own questions off its servers

In every row of that table, your chat sits on the AI company's servers. On a personal account, it sits there under terms you accepted with a click and can't negotiate. A business contract gets better terms for the organization that signs it, and the chats still sit on someone else's servers.

Morse is for your own questions, on your own device. Work material belongs in the tools your employer approves, under its rules. For the rest, Morse is built so your chat history doesn't wait on a server under someone else's terms:

• Private, for the questions you'd keep from any AI company. Every new chat opens here. The models are open-weight ones such as DeepSeek, Kimi, and GLM, along with Grok, and Morse has zero data retention agreements with its inference, search, and other providers, so they keep nothing. When a provider confirms zero retention reply by reply, Morse verifies it and turns away any reply that comes without it.
• Anonymous, for reaching GPT or Claude as Morse rather than as you. Before the message goes out, our proxy takes off your account, email, and IP address, and the provider receives it on Morse's account, without knowing it's you. Under its own policy, the provider may hold the message for a limited time to check for abuse, so write the question without your name or your employer's.
• E2EE, for the things you'd show no one. Before anything is sent, your browser verifies the enclave's hardware attestation and seals the message to that one enclave, so Morse's servers pass along ciphertext they can't open. Third-party providers run the enclaves, not Morse, and the protection holds as long as the enclave hardware and attestation work as designed. E2EE comes with Pro; here's how end-to-end encrypted AI chat works.

In every mode, there's no training switch to find, because Morse doesn't train AI on your content, sell your data, or show ads (Morse privacy policy). Your chat history is stored only in your browser, encrypted, not on Morse's servers, and our servers don't keep your messages or log what you write. History locks when you sign out and takes two keys to open, one of which never leaves your device. Because we don't store chats, we usually have very little to give authorities. Clearing your browser or deleting your account removes your chats for good, and Morse can't recover them, so keep a copy of anything you need.

Send a first message from the Morse home page on your own device without an account, or start free with no card needed. End-to-end encrypted chats come with Pro, at $19 a month.