PAIR Finance brings AI debt collection to Norway, but can it prove what the system decided?

When the postcard arrives

PAIR Finance has brought its AI-driven debt collection platform to Norway ^[1]. The company, which already operates in Germany and Austria, says its AI decides when to contact debtors, what to say and which channel to use. It claims recovery rates 15% higher than traditional methods.

The Norwegian launch matters because it shows how automated decision-making is spreading into areas that directly affect people's finances and reputations. When an AI decides whether to send a payment reminder by post, email or SMS, and when to escalate a case, those decisions shape what happens next. And under EU rules coming into force in December 2027, they will also need to be logged.

Think of it like a postcard. When you post a letter, the postmark shows when and where it entered the system. Nobody can change that stamp afterwards. Debt collection decisions are a bit like that: they happen on a specific day, and months or years later someone may ask what the system decided and why. The difference is that a postcard's stamp is printed by a machine at the sorting office, while an AI's log is kept by the company running the AI.

What the AI Act asks for

The AI Act classes systems for credit assessment of individuals as high-risk ^[2]. That includes decisions about whether someone gets a loan, but also decisions about collecting debts, since they affect a person's financial standing and can lead to legal action.

Article 12 of the AI Act says high-risk AI systems must technically allow for the automatic recording of events over their lifetime ^[3]. The rule applies to Annex III systems, which include credit assessment, from 2 December 2027. It says what to log: enough to trace the system's operation and support post-market monitoring. It doesn't say how to protect the logs once they exist.

Article 19 adds that providers must keep the logs they control for at least six months, unless another law requires longer ^[3]. For users of the system, Article 26 says the same thing.

The question nobody asked yet

Here's the part that tends to get less attention. When a debtor disputes a decision, the company's log is the main evidence of what the AI decided and when. But the company controls the log. It's kept on the company's systems, accessible to the company's administrators, and backed up by the company's infrastructure.

That's fine until someone questions it. Then the company is in the position of saying: trust us, we didn't change anything. That works until it doesn't.

In Norway, the Debt Collection Act (inkassoloven) already sets rules about how debts must be collected, including how and when debtors must be contacted. PAIR Finance says its AI optimises those decisions by analysing data and choosing the right moment and channel. Fair enough. The tricky bit is showing, later, that the decision the AI made on 14 March is the one in the log you're looking at in October.

What a sealed log does

This is where a digital witness fits. The thing to seal is each decision record as it's written: the debtor reference, the AI model version, the decision (for example, send reminder by email), the timestamp and, if there is one, the staff member who reviewed it. Sealing takes a fingerprint of the record, not the record itself, so the decision data never leaves PAIR Finance's systems.

When a decision is questioned months later, anyone the company shares the record with can check that it's exactly what was written on the day. Change a single character in the log and the fingerprint won't match. The check doesn't require trust, because the sealed fingerprint is on a chain where even the company that sealed it can't rewrite history quietly.

A witness like this doesn't show that the decision was fair, lawful or correct. It shows that the log is the one that was written at the time. That's a smaller claim, but it's the one that matters when someone says: I don't believe that's what your system decided.

How it would work

1. PAIR Finance's system writes a decision to the log.
2. The system sends a fingerprint of that record to WitniumChain through the REST API, the SDK or the MCP server, and gets a witness ID back.
3. Months later, if the decision is disputed, the company shares the log record and the witness ID. The other party calculates the fingerprint again and checks it against the sealed one. If it matches, the record hasn't changed.

What this means for you

If you run a business that uses AI to make decisions about individuals, here are three questions worth asking now:

• Could you show, six months from now, what your AI decided on a given day, in a way that doesn't rely on someone taking your word for it?
• Who in your organisation has access to the logs, and could they edit them, even by mistake?
• If a customer, a supervisor or a court asked to see the decision record, would the record itself be enough, or would you end up in a conversation about whether it's been changed?

The AI Act's logging requirement applies from 2 December 2027 for Annex III high-risk systems. That's less than three years away. Getting the logs right is table stakes. Making them hold up when questioned is the part that takes a bit more thought.

What to watch

2 December 2027 is when the high-risk requirements, including automatic logging, apply to Annex III systems ^[2]. Credit assessment of individuals is one of them. If you're building or buying AI for decisions about people, that date is worth marking.

Norway isn't in the EU, but it tends to follow EU rules closely through the EEA agreement. The AI Act is expected to be incorporated into EEA law, which would make it apply in Norway as well. Worth checking how that process is going if you operate there.