Last updated: August 9, 2026 · By Vishal Swami, Founder & Lead AI Reviewer, AISagely
Denmark's Ministry of Education announced on August 7, 2026 that upper-secondary students must now defend their take-home written assignments out loud, in person, in front of a teacher. The rule targets the roughly 9,000 students a year who complete the SSO, Denmark's major written assignment for the two-year HF program, and it took effect immediately.
Short answer: Denmark now requires oral defenses of all take-home written exams for its roughly 9,000 HF-program students, effective from the 2026 school term. Schools must also add screen-monitoring software and content-filtering firewalls during in-school tests. Education Minister Magnus Heunicke announced the emergency measures on August 7, 2026, citing widespread AI-assisted cheating in upper-secondary schools.
I spend most of my working hours testing AI tools, including the ones students use to write essays in ten minutes flat. Denmark's fix is interesting because it doesn't try to out-tech the problem with a detector. It just asks a student to explain, out loud, what they claim to have written. That's a low-cost move any teacher could copy, and it's worth understanding even if you've never set foot in a Danish classroom.
What Denmark's new rule actually requires
Minister Heunicke put it bluntly: "Unfortunately, we have a problem with students using AI to cheat in upper secondary school. Action is needed now, and we are starting with these three initiatives," he said, according to Euronews’ report on the announcement. The three initiatives are the oral defense requirement, mandatory screen-monitoring during written tests done at school, and firewalls that filter what students can access online during class and exams.
The oral defense rule is aimed squarely at the SSO, or større skriftlige opgave, the major written assignment roughly 9,000 HF students complete each year. From now on, a student has to sit down with a teacher and explain their argument, their sources, and their reasoning, without the paper doing the talking for them. TechRepublic’s coverage reports that schools are also being pushed to move more written work back into the classroom, under supervision, rather than sending it home where a chatbot is one tab away. Students must now also declare when they used AI on a major written assignment, and preparation for oral exams has to happen without AI access.
Anders Frikke of the Danish Association of Upper Secondary School Teachers welcomed the speed of the response. Oscar Tønsberg Hoffmann, who represents the country's upper-secondary students, raised a fair concern: pushing more writing into the school day only helps if it's organized well, not just moved from a bedroom desk to a classroom desk. Denmark had actually been leaning the other way before this. Euronews notes the country had experimented with letting students use AI during English oral tests. This rule is a sharp reversal, driven by how fast that experiment ran into cheating at scale.
What you'll need to run something like this
You don't need Denmark's ministry to try this. Any teacher can add a short oral component to a written assignment with nothing more than five to ten minutes per student and a handful of questions about the work itself. What you do need is a written product that's substantial enough to defend, an essay, a lab report, a research paper, not a one-paragraph response. You also need a private minute or two per student, since the point is hearing them explain their own reasoning without a classmate feeding them an answer.
Step-by-step: adding an oral defense to written work
1. Keep the writing assignment exactly as it is
Don't redesign the assignment around the defense. The paper itself stays the deliverable. The oral component is a check on it, not a replacement for it.
2. Ask for the paper before the conversation
Collect the written work first, and skim it before the student sits down. You're listening for whether their spoken explanation matches what's on the page, not grading a second essay.
3. Ask about choices, not content
Generic recall questions ("what does your third paragraph say?") are easy to fake with five minutes of skimming. Questions about choices are harder: why this source over another, what a specific counterargument was and why they didn't use it, what they'd change with another week.
4. Time-box it
Two to five minutes per student is enough for most assignments. Denmark's SSO defenses reportedly run longer given the assignment's size, but for a normal essay, you're listening for confidence and specifics, not conducting a full thesis defense.
5. Score the gap, not just the answer
A student who wrote the paper themselves will stumble on some questions and nail others, the normal texture of real knowledge. A student reciting a chatbot's summary of its own output tends to answer everything at the same generic depth. That flatness is the tell.
6. Give a real second chance
Denmark's plan lets students contest a flagged grade. Build in something similar. A nervous 16-year-old bombing a two-minute oral check isn't the same as a student who can't explain a single claim in their own paper.
Example questions you can copy
Swap in your subject, but the shape of these works for almost any written assignment:
- "You cited three sources. Walk me through why you picked that third one over an obvious alternative."
- "What's the strongest argument against your own thesis, and why didn't you address it?"
- "If you had one more week, what's the first thing you'd change about this paper?"
- "Explain your second paragraph to me as if I hadn't read it."
In my testing, I fed the first two questions into ChatGPT along with a sample five-paragraph essay and asked it to draft answers a student could memorize in advance. It produced clean, confident-sounding responses in under a minute. But they stayed generic on the specific source comparison, and completely dodged naming a real counterargument until I pushed it twice. A student who'd actually read their own sources would answer the second question in one sentence. A student running on a memorized script tends to circle it.
Common mistakes to avoid
The biggest mistake I'd flag, based on how similar tricks have played out elsewhere, is picking questions that are too easy to prep for. If your questions map one-to-one onto the essay's headings, a student can just memorize a summary of each section without ever having engaged with the argument. Go for the parts that require reasoning in the moment.
Second, don't skip the paper-first step. Judging a defense without having read the written work means you're grading confidence, not accuracy, and confident delivery is exactly what a memorized script produces.
Third, resist grading purely on nerves. Some of the strongest student writers freeze up the moment a teacher asks them a direct question, especially the first time a class runs this. A rubric that separates "can't explain the work" from "explains it badly under pressure" matters here.
Fourth, don't treat this as a one-time fix. The professor behind the invisible-text prompt trap that caught 32 of 35 students cheating with AI found the same thing Denmark is likely to run into: once a detection method gets famous, students adapt to it. Rotate your approach.
Oral defense vs. other AI-cheating countermeasures
| Method | Cost | Setup time | What it catches | What it doesn't prove |
|---|---|---|---|---|
| Oral defense (Denmark's approach) | Free | 5–10 minutes per student | Whether a student understands their own written argument | That every sentence in the paper is 100% student-written |
| AI detection software (e.g., Turnitin's AI indicator) | Usually included in institutional licenses | None — runs on submission | Statistical patterns common in AI-generated text | That a specific student didn't write a specific passage |
| Screen monitoring during in-class writing | Requires school-issued software | Ongoing, per exam | Tab-switching to a chatbot during a supervised test | Cheating done before or after the monitored window |
| Invisible-text prompt trap | Free | About 2 minutes per assignment | A student who copy-pastes a question into a chatbot unread | Cheating by a student who edits the AI output before submitting |
Tools that make this easier
None of this needs new software, but a few resources help the surrounding workflow. If you're a teacher building assignments and rubrics around this, my best AI tool for teachers guide covers what's actually worth using for the parts of the job that aren't the defense itself, like rubrics and differentiated questions. My broader best AI tool for education roundup compares tools built for classroom use more generally. On the student side, if you want to use AI honestly while still being able to defend your own work, my guide to using ChatGPT to write an essay without plagiarizing covers where that line sits in practice. If you're trying to spot AI-written passages before a defense even happens, how to spot AI writing walks through the same tells I use. And if you're a student figuring out which AI tools are actually worth using for real coursework instead of shortcuts, my AI tools for students guide has the honest list.
My take
I like this rule more than I expected to. AI detection software flags plenty of human writing as machine-generated, and every new detector gets reverse-engineered within a semester. An oral defense doesn't have that failure mode. It doesn't care how the paper was produced. It only checks whether the student sitting in front of you actually understands what they turned in. That's a bar honest AI use clears easily, brainstorming with a chatbot and then rewriting in your own words still leaves you able to explain your choices, and a bar copy-paste cheating doesn't. The tradeoff is real: five to ten minutes per student adds up fast in a class of thirty, and Denmark's own students are right to worry that the extra classroom writing time needs to be worth something. But as a check that's cheap, hard to game, and doesn't require trusting a detection algorithm, it's one of the more sensible responses I've seen this year.
Frequently Asked Questions
Does Denmark's oral defense rule apply to every student, or just some?
It applies specifically to the roughly 9,000 students a year in Denmark's two-year HF (Higher Preparatory Examination) program who complete the SSO, the country's major written assignment. The broader push toward more in-school, supervised writing applies more widely across upper-secondary schools.
When did this rule take effect?
Education Minister Magnus Heunicke announced the measures on August 7, 2026, and they took effect immediately, ahead of the new school term.
Is an oral defense the same as an AI detector?
No. AI detection software looks for statistical patterns common in machine-generated text. An oral defense checks whether a student can explain the argument, sources, and reasoning behind their own paper, which doesn't depend on catching a particular writing style.
Can a teacher outside Denmark use this method?
Yes. It doesn't require any special software or institutional policy, just a few minutes per student and a set of questions that ask about choices rather than content recall. See the step-by-step section above for how to set one up.
What other measures is Denmark adding alongside oral defenses?
Schools are also being required to use screen-monitoring tools during in-school written tests and firewalls that filter what students can access during class and exams, along with encouragement to move more written work into supervised classroom time.