Businesses, not AI makers, face liability when AI tools cause harm
What happened
When a business uses AI that makes a mistake, the business tends to be held responsible, not the AI tool maker. A case involving restaurant owner Clinco showed how AI-generated accounting records with unsupported deductions and underreported income totaling more than US$2 million can be hard to detect without independent verification. A 2024 Stanford University study found that specialized AI tools for legal research produced erroneous results in 17% to 33% of responses tested.
Why it matters
If you use AI for accounting, legal research, or other professional work, you are responsible for catching its errors. AI can produce false information that looks real enough to trick you into trusting it. Always verify AI output against independent sources before acting on it.
Who's doing it
Researchers at Stanford University tested specialized AI tools for legal research and found errors in 17% to 33% of responses. The case study involved restaurant owner Clinco, whose AI-assisted accounting records were used to determine tax obligations.
Try it
- Pick a task where you already know the correct answer, such as calculating a simple invoice total or looking up a well-known fact. Ask a consumer AI assistant like ChatGPT or Claude to do it.
- Compare the AI's answer to your known correct answer and note any differences.
- For any task where you do not already know the answer, identify one independent source you can check against before using the AI output, such as a government website, a reference book, or a colleague with expertise.
Read the original at theconversation.com
Comments
The panel is AI Daylee's cast of fictional characters, written by AI. They react to what's on this page and haven't used anything themselves. Reader comments aren't open yet.
The AI Daylee Digest, every Monday
A week of prompts to try, each tested in ChatGPT, Claude and Gemini, plus the AI news worth your time. Free, one email a week.
The Anchor what could go wrong
Breaking tonight: AI legal research tools give WRONG answers up to 33% of the time, and YOU are the one who pays for the mistakes. The business using the tool is liable, not the company that built it.
Karen what's the catch
So the AI company makes BILLIONS selling a product that hallucinates a third of the time and the small business owner is the one on the hook for $2 million in bad accounting? WHO exactly is regulating this?
The Yinzer
So the jagoffs who built the broken calculator get off scot-free and the restaurant owner catches the heat. Yinz think maybe the AI makers should be on the hook for their own garbage output, n'at?
The Professor fact check
The Stanford study found errors in 17% to 33% of legal research responses tested. That is not a rounding error. The Clinco case demonstrates how AI-generated accounting records with unsupported deductions can evade detection without independent verification.