Swedish doctors reject AI journal tool as 'excessive faith in technology'

What happened

Swedish healthcare is rolling out an AI tool that listens to doctor–patient conversations and writes journal entries. The plan, according to Läkartidningen, is for 80% of hospitals and primary care centres to adopt it by autumn 2025. ^[1]

Doctors are not impressed. In comments reported by Läkartidningen, they describe the push as 'excessive faith in technology', and some say it puts the technology before the professional responsibility of documenting care. ^[1]

The case raises a few practical questions. One is about responsibility: who stands behind an entry the AI wrote? Another is about evidence: if an entry is questioned later, how do you show it is the one written on the day, and that nobody has quietly edited it since?

Who is responsible for an AI-generated note?

Under Swedish patient data law, journal entries must be signed by the person responsible for them. ^[2] The law does not say an AI cannot help write the entry, but it does say a person has to sign it.

So if an AI listens to a consultation and produces a draft, the doctor still has to read it, decide whether it is accurate and sign it. If it is wrong, the signature is still the doctor's.

The AI Act adds another layer. When a high-risk AI system is used in a workplace, the employer must inform employees and their trade union representatives before it is put into use. ^[3] That requirement applies from 2 December 2027. Whether this journaling tool counts as high-risk depends on how it is used and what decisions it informs, but the rule is there.

The upshot is that responsibility does not disappear just because the technology wrote the first draft. Think of it like a spell checker: it can suggest, but it cannot sign.

The evidence question

Journal entries are evidence. When a patient asks what was decided and why, the journal is meant to answer. If someone questions whether an entry was changed, the record system needs to show that it was not.

Most journal systems keep audit logs of who opened a record and when, but that is not quite the same thing. An audit log shows access; it does not show that the entry itself is unchanged. And the people who could change an entry are often the same people who could, in theory, change the log.

Here is where a digital witness fits. The thing to seal is the fingerprint of each signed entry, not the content. The entry stays in the record system; the fingerprint is sealed on a chain where a later change would show. If an entry is questioned, anyone with the original can calculate the fingerprint again and check it against the sealed one. Change a single character and it will not match.

What this means for you

If you work with journal systems, or any system where AI helps write records that matter, ask:

• Who is responsible for checking and signing what the AI writes?
• Can you show, months later, that a signed entry has not changed?
• If the system is classed as high-risk under the AI Act, have you told employees and their representatives?

For vendors, the rule is simpler. If your software helps write journal entries, make sure the person using it knows they are still the one who signs.

What to watch

• 2 December 2027: the high-risk requirements in Annex III of the AI Act apply, including the duty to inform employees before deployment. ^[3]
• 2 August 2028: the high-risk requirements in Annex I apply. ^[4]

How it would work

1. When a doctor signs an entry, the system calculates a fingerprint and seals it with WitniumChain, through the REST API, the SDK or the MCP server. The entry itself stays in the record system.
2. If the entry is questioned, the system calculates the fingerprint again. If it matches the sealed one, the entry has not changed.
3. Change one character and the fingerprints will not match, even if the person who changed it has administrator rights.

Read the docs for how to integrate it. The first 2,000 witnesses a month are free.