How Skolkoll uses AI
Where AI is actually used on Skolkoll, what controls are in place — and, just as importantly, what the AI does not do. We describe only what is actually built.
Skolkoll uses AI in a few, well-defined places — chiefly the assistantKollen, but also to stylise school images, to draft journalist-reply emails and as internal drafting aids behind the scenes. Two things hold everywhere, and they are the guarantee that matters most: the AI never sets any published numbers, and nothing an AI creates is published or sent on without a human approving it — whether a school image, a journalist reply or data from a document. The one exception is Kollen itself: an AI assistant you actively choose to talk to, clearly labelled as AI, which only answers your own question back to you and never invents numbers. This page lists every place openly, what safeguards exist and where the limits are. We would rather understate than overstate: a control that is not built is not listed here, and a place where we use AI is not hidden away.
1. Where we use AI
Kollen — the assistant that answers questions about schools
Kollen is an AI-based chat that answers questions about a school, or about school statistics in general. When you ask a question, it is sent — together with the relevant page context — toAnthropic (USA) via their Claude API, which generates the answer. The model that answers isClaude (Sonnet). Kollen requires you to actively consent before the chat starts, and the answers are built on the verified data Skolkoll has already calculated — not on anything the AI makes up.
Anthropic processes your messages as a subprocessor and may retain the content for up to 30 days for security purposes. Skolkoll does not store the message content permanently. The full data-protection details are in theprivacy policy and inData protection and subprocessors.
School images — an AI-stylised illustration, always human-reviewed
The image at the top of a school page is not a photograph of the school but anAI-stylised illustration in a chalkboard style. The source — often a façade photo — is sent to an image model at OpenAI that transforms it into the stylised version; the image is processed under OpenAI's own safety and content policies. But no AI-stylised image is published automatically: our editors review and approve every image before it is shown. The source is either submitted by the school (with consent) or a licensed, credited photo — for example from Wikimedia Commons — and the provenance is always shown in the image's (i) panel. The illustration is just an illustration — itnever affects grades, numbers or ranking. More on the data protection is inData protection and subprocessors.
Email — template outreach without AI, but journalist replies a human sends
There are two different paths here, and the difference matters. Pilot outreach to schools and providers is built from a deterministic template in which predefined{{variable}} fields are filled in with data from our own registers — no language model is involved, and every such message is predictable and auditable. However, when a journalistgets in touch, AI may help draft a reply. That draft is never automatic — a human reads, edits and sends every such email. The AI may compose text on one path, but never presses "send" itself.
Internal drafting aids — human-gated, never reach the outside world on their own
In a couple of places behind the scenes, Skolkoll uses AI as a working tool, never as something that publishes or sends on its own. AI may extract data from official documents into a review queue where a human approves before anything is used — automatic publishing is disabled. We are also piloting AI-assisted triage of support tickets, where a suggested reply only lands as aninternal note and is never sent automatically. In every case the AI is a suggestion on a human's desk — none of it reaches a family, a school or a published number without a human having made the decision.
AI never sets any published numbers
All statistics on Skolkoll — merit scores, SALSA results, eligibility rates, the Skolkoll score and every other key metric — are calculated deterministically from source data from agencies such as Skolverket, SCB and Kolada. The AI never produces, alters or selects any of these metrics. The one case where an AI reads numbers out of official documents (see the internal drafting aids above) lands in a review queue where a human decides — nothing is published automatically. When Kollen mentions a number, it comes from the verified source data, not from the model. How the numbers are calculated is documented openly on themethod page.
2. What controls are in place
Several layers of safeguards surround Kollen. Some are hard guarantees; others are deliberately described as "best-effort" because they can let something through in edge cases — we would rather be honest about the limits than promise more than the code delivers.
Grounding in verified data
On a school page, the server injects the school's verified Skolkoll data as context before the question reaches the model. The system prompt explicitly tells the model to answer "I don't have data for that right now"rather than guess, and to never invent statistics or school names that are not in its context. The source data is the source of truth; the model is expected to stay within it.
Input sanitisation
All fields in the school context are sanitised before being built into the prompt, so that content in the data cannot be interpreted as instructions to the model.
Scoped to school questions
The prompt includes an instruction to ignore attempts to inject new instructions and to answer only questions about schools and school statistics. This is a mitigation, not a guarantee — it reduces the risk of manipulation but cannot rule it out entirely.
Best-effort topic screening
Before a question is answered, it is first classified by a faster model (Claude Haiku) as on-topic or off-topic. This is a best-effort control: if the classification fails, the question is allowed through rather than wrongly blocked (it "fails open"). We therefore never describe it as a safeguard that blocks everything inappropriate. Separately, a circuit breaker can temporarily disable the whole AI chat if the main model fails repeatedly — an availability safeguard, not part of the topic screening.
Request integrity (HMAC signing)
The assistant's messages are signed with HMAC and re-verified on subsequent requests. If a message has been tampered with, the request is rejected. This means the conversation history cannot be forged in transit.
Rate limits
Requests are limited in several ways: a short-term per-IP burst limit, a daily limit and — for signed-in users — a monthly quota. This protects the service against overload and abuse.
Audit logging with a hashed IP
For every AI request, a pseudonymised record is written to the ai-audit-log collection: a SHA-256 hash of your IP address (not the full address), the length of the question and answer (not the content), the school-context code, status and a timestamp. Records are tagged with a TTL of90 days and deleted thereafter. No message content is logged.
An always-visible "AI can be wrong" disclaimer
Every answer from Kollen is automatically appended on the server with a source-and-fallibility disclaimer —"AI-generated summary based on data from Skolverket, SCB and Kolada via Skolkoll.se. Kollen may make mistakes — verify important information." In addition, the chat interface shows a permanent disclaimer carrying the same call to verify:
"Kollen drivs av AI (Claude, Anthropic). Kan göra fel — verifiera viktig information."(Kollen is powered by AI (Claude, Anthropic). It can make mistakes — verify important information.)
The disclaimer is deliberate and permanent. The numeric reasonableness check we run internally is alog — it does not correct or block answers. So the disclaimer stands as the visible safeguard: the source data is the source of truth, the model is grounded on it, and you are always asked to verify important information.
Explicit, revocable consent
Kollen does not start until you have actively consented. You can revoke consent at any time via the "Revoke AI consent" button in the chat. The conversation itself is stored only for the browser session and disappears when you close the tab; the consent remains until you actively revoke it.
3. What the AI does not do
- It sets and alters no published statistic. Every key metric on Skolkoll is calculated deterministically from the agencies' source data. The AI never produces, alters or selects any metric — it can only read and describe the data the server has already calculated.
- It sends and publishes nothing on its own. Kollen cannot send email, store data or perform actions. Where AI helps compose something that reaches the outside world — journalist replies, support drafts, stylised school images or data extracted from documents — a human always reads, approves and sends. No AI text and no AI image leaves Skolkoll without a human having made the decision.
- It makes no automated decisions about individuals. Skolkoll does not engage in automated decision-making with legal effect for individuals. Kollen answers questions about public school statistics — it does not assess, rank or decide about people. This is consistent with our record of processing activities (ROPA).
We keep this page honest about the limits: the topic screening fails open, and the instruction to answer only school questions is a mitigation rather than a guarantee. If a safeguard is improved or added, this page is updated accordingly.