IMY puts public sector AI under the microscope in 2026

A priority, not a surprise

IMY, Sweden's data protection authority, has announced that supervision of AI use in the public sector will be one of its priority areas for 2026 ^[1]. The authority will support public bodies in deploying AI compliantly and check that they can demonstrate what their systems did when someone asks.

It's a bit like when the tax agency announces it will focus on a particular sector. The rules were always there, but now everyone knows where the torch is pointed.

Public authorities already process large amounts of personal data, often in decisions that directly affect people's lives: social services, healthcare, education and law enforcement. Adding AI to that picture raises the stakes. The question is not whether AI can be used, but whether it can be used in a way that people can trust and that holds up to scrutiny.

What the law says

Several uses of AI in the public sector are classified as high-risk under Annex III of the AI Act. That includes AI for biometric identification, law enforcement, migration and asylum, and administration of justice ^[2]. For those systems, high-risk requirements apply from 2 December 2027, after the Digital Omnibus postponement ^[3].

Those requirements include risk management, data governance, technical documentation, automatic logging under Article 12, human oversight and quality management, and a conformity assessment before the system is placed on the market ^[2].

Article 12 says high-risk AI systems must technically allow for the automatic recording of events over their lifetime ^[4]. What it does not say is how to protect those logs, but their evidential value is a reason to do so.

Article 26 sets out duties for deployers. Public authorities using high-risk AI must follow the instructions, maintain human oversight, monitor the system in operation and inform people affected by decisions ^[5]. Article 27 adds a specific obligation: public actors, and private actors performing public tasks, must carry out a fundamental rights impact assessment before deploying a high-risk system ^[6].

Under the GDPR, Article 22 already limits automated individual decision-making that has legal or similarly significant effects on people, unless it is necessary for a contract, authorised by law or based on explicit consent ^[7]. Even when allowed, there must be a right to human intervention, to express one's view and to contest the decision.

What this means for you

If you work in or with a public authority that uses AI, or plans to, here are the questions worth asking now:

• Does the AI system fall into any of the Annex III categories, making it high-risk?
• If it does, can you show what the system logged on a given day, and that the log has not changed since?
• Has a fundamental rights impact assessment been done before deployment?
• Are people informed when a decision about them involves AI, and do they have a way to contest it?
• Is there meaningful human oversight, or is the AI making the call on its own?

Those are not hypothetical questions. They are the ones IMY, and ultimately a court, will ask if something goes wrong.

Logs that hold up

A log kept by the same organisation that could edit it is easy to challenge. The high-risk requirements do not prescribe how to protect logs, but the supervision priority suggests IMY will want to see that public bodies have thought it through.

Here is where a digital witness fits. The thing to seal is each log record as it is written: the person affected, the system version, the decision and the human reviewer who signed off. Sealing takes a fingerprint of the record, not the record itself, so the log stays in the authority's own systems. If a decision is challenged six months later, anyone the authority shares the record with can check that it is exactly what was written at the time.

Auditors, supervisors and, if it comes to it, a court can all do that check. The record does not prove the decision was right or lawful, but it does show it has not been quietly rewritten.

Your system sends a fingerprint to WitniumChain through the REST API, the SDK or the MCP server, and gets a witness ID back. To check, the fingerprint is calculated again and compared with the sealed one. The content never leaves your systems. The first 2,000 witnesses a month are free, and the documentation is public.

A digital witness does not replace access control, monitoring or backups. It gives you one thing: a way to show that a log entry has not changed, even when you are the one keeping it.

What to watch

High-risk requirements for Annex III systems, including the logging duty in Article 12, apply from 2 December 2027 ^[3]. Public sector deployers of high-risk AI must conduct a fundamental rights impact assessment under Article 27 before deployment ^[6].

IMY's supervision priority for 2026 means guidance and checks are already on their way. For public authorities planning AI deployments, the time to ask these questions is before the system goes live, not after someone disputes a decision.