Case Studies · Ungoverned AI
Real leaks, lawsuits, and errors - and what would have stopped them.
A library of incidents from companies around the world.
Poland's supreme administrative court caught a lawyer citing AI-fabricated rulings
Three of the court decisions in the filing didn't exist. The court noticed before anyone else did.
Read the caseAn undisclosed AI-written government report cost an official her job in 24 hours
It wasn't the AI use itself that ended her position. It was that nobody knew until critics found the tells.
Read the caseDeloitte refunded part of a government contract after AI-assisted errors
A single unverified citation was enough to put an entire deliverable in question.
Read the caseNew York now requires a bias audit before AI can screen a candidate
The regulation didn't ban AI in hiring. It required proof it isn't discriminating.
Read the caseA road-maintenance bidder lost a 15.5M PLN contract to AI-invented tax rulings
Nobody checked the AI's citations before they went into a winning bid - until a competitor did.
Read the casePoland's data regulator went public asking for AI hiring safeguards before the first fine
The regulator didn't wait for a scandal - it named the risk and asked for rules in advance.
Read the case101,000 stolen ChatGPT logins showed how much shadow AI use is actually happening
Employees were pasting company data into personal ChatGPT accounts. Malware found the accounts before anyone else noticed the pattern.
Read the caseA law firm used ChatGPT to justify its fees - and the judge wasn't convinced
The AI wasn't wrong, exactly. It just wasn't evidence.
Read the caseAir Canada was held liable for what its chatbot told a customer
The chatbot was wrong. The company was still on the hook.
Read the caseSamsung employees pasted confidential source code into ChatGPT
Engineers used a public chatbot for day-to-day work, and sensitive code left the building without anyone deciding it should.
Read the caseApple restricted ChatGPT over confidentiality concerns
The risk wasn't the AI itself - it was employees quietly moving sensitive work into a tool outside the company's control.
Read the caseAmazon warned staff after employees leaned on ChatGPT for work
Employees wanted to use AI. The company wanted that too - just not without knowing where the data was going.
Read the caseLawyers cited fabricated cases ChatGPT invented
AI generated something that looked exactly like a real legal citation. It wasn't.
Read the casePoland's data regulator opened a formal case against OpenAI over ChatGPT
A researcher asked ChatGPT for facts about himself. It made some up - and there was no way to get them corrected.
Read the caseClearview AI was fined €20 million for scraping faces without a legal basis
The technology wasn't the violation. Not knowing what data you're processing - and under what legal basis - was.
Read the caseAn algorithm sorted unemployed Poles into tiers, and a court found it unconstitutional
A scoring system decided how much help someone got looking for work. Nobody could see how it scored them.
Read the caseAmazon's recruiting AI learned to downgrade resumes from women
The model didn't invent the bias. It learned it from the company's own hiring history.
Read the caseWhere does your organization sit in these stories?
Most companies are closer to one of these than they'd like to admit.
Book a Consultation