Olympic Ice Dancers Hit Plagiarism Double-Fault With AI Music Generator
Everyone keeps telling us AI will enhance human creativity, not replace it. That it's a tool, like Photoshop or Auto-Tune. Well, Katerina and Daniel Mrazek just proved that conventional wisdom wrong at the 2026 Winter Olympics.
These Czech ice dancing siblings performed their rhythm dance on February 9th using a track literally labeled "One Two by AI (of 90s style Bon Jovi)." Not exactly hiding it. But here's where it gets spicy: this was their second AI music attempt after the first one plagiarized New Radicals' "You Get What You Give" bar-for-bar in an AC/DC style.
<> "LLMs sometimes spit out straight-up plagiarism" - and Olympic ice dancing just became the poster child for this problem./>
The ISU approved both tracks. Think about that for a second. The governing body of ice dancing rubber-stamped what NBC commentators casually mentioned during live broadcast as AI-generated music, complete with lyrics that echo Bon Jovi's "Raise Your Hands" ("raise your hands, set the night on fire").
Sports analysts called it "stupid." BroBible highlighted the plagiarism scandal. Social media exploded. And honestly? They're right to be pissed.
Ice dancing is 90% music and choreography. It's not figure skating with jumps and technical elements dominating scores. The artistry is the sport. When journalist Shana Bartels noted the Bon Jovi similarities back in November 2025, she was documenting the death of authenticity in real-time.
The Technical Trainwreck
From a developer perspective, this is LLM hallucination and training data regurgitation at its worst. The model didn't "create" anything - it remixed copyrighted lyrics it memorized during training. That's not intelligence. That's a very expensive copy-paste machine.
Want to build music AI? Here's what the Mrazek incident teaches us:
1. Implement plagiarism detection before output
2. Use RAG with licensed sources instead of raw generation
3. Add watermarking for provenance tracking
4. Fine-tune with deduplication to avoid verbatim reproduction
The vocal synthesis mimicking Bon Jovi's timbre raises deepfake audio concerns too. We're not just talking copyright violation - we're talking identity theft.
The Elephant in the Room
Nobody wants to admit this, but the Mrazeks aren't positioned to medal. They're Olympic debutants who probably couldn't afford custom music composition. AI seemed like a cheap solution to meet the ISU's "90s rock style" requirement for rhythm dance.
But that's exactly the problem. AI democratizing creativity sounds great until you realize it's democratizing plagiarism. The barrier to entry wasn't just cost - it was the creative process itself. Finding the right song, securing rights, working with choreographers to match music to movement. That friction forced intentionality.
Now? Generate, approve, perform, controversy.
The ISU's regulatory gap is glaring. They verified the paperwork but not the originality. Sports organizations worldwide are watching this unfold, realizing they need auditable AI pipelines for competitive use.
The real tragedy? Other ice dancing teams spent months curating perfect musical selections, negotiating licenses, crafting narratives through sound. The Mrazeks used a prompt.
This isn't about being anti-AI. It's about AI companies training on copyrighted data without consent, then selling access to laundered plagiarism. The model that generated "One Two by AI" was fed New Radicals and Bon Jovi tracks, then spit out barely-disguised copies.
TechCrunch called it a "depressing symbol" of AI reliability issues. I'd go further: it's a symbol of an industry that prioritizes impressive demos over ethical training data.
The 2026 Olympics will be remembered for many things. Czech ice dancers accidentally proving that LLMs are sophisticated plagiarism engines probably wasn't on anyone's bingo card.
