AI Literacy Framework · Part 6 of 7
Transparency and accountability
Own every claim. Disclose how AI was used.
Transparency and accountability diligence
This is the diligence you apply at the end of the workflow, before anything leaves your hands.
Accountability
You own the output and answer for every claim in it.
“The AI wrote it” is not a defence.
Transparency
Tell the people who need to know how AI was used: supervisor, partner, participants, instructor.
Match the detail to how much AI shaped the work.
You are the final checkpoint.
Here’s a test. Could you explain and defend every sentence if someone asked, without the AI open? If not, it isn’t ready to leave your hands.
Diligence failures
Five real cases from law, education and business. As you read, ask of each one: was it an accountability failure (nobody checked), a transparency failure (nobody said), or both?
Law
Zhang v. Chen (BC, 2024)
- What happened — A Vancouver lawyer filed a court application that cited two cases ChatGPT had invented.
- How it came out — The other side’s lawyers went looking for the cases and couldn’t find them. She had not checked them herself.
- The court’s finding — She must personally pay the costs her error caused the other side, and review her other files for AI-generated citations.
She was open about it once caught, which is part of why the penalty was limited.
Zhang v. Chen, 2024 BCSC 285, BC Supreme Court, February 2024. CBC News
Law
Ko v. Li (Ontario, 2025–26)
- What happened — A Toronto lawyer filed a court document drafted with ChatGPT. The cases it cited were non-existent or irrelevant.
- What she told the court — That she hadn’t used ChatGPT herself and someone else had prepared the document. Months later she admitted she wrote it alone.
- The outcome — A six-month suspension, $10,000 in costs, and a referral to the Attorney General for possible criminal contempt.
The unchecked citations were a mistake she had nearly recovered from. Hiding how the work was made is what led to the six-month suspension.
Ko v. Li, 2025 ONSC 2766 and 2025 ONSC 6785; Law Society Tribunal order, 16 July 2026. Law Times · WeirFoulds LLP
Education
Dartmouth’s provost (2026)
- What happened — The college’s top academic officer published a newspaper op-ed about AI and student learning. He had used ChatGPT on it and didn’t say so.
- His response — He said the ideas were his own and AI only refined the wording. He later said he should have disclosed it.
- The outcome — A faculty-led review found he had used AI to meaningfully alter his writing and hadn’t been forthright about it, short of the college’s academic integrity standards. He resigned in October.
He called it help with wording. The review found AI had shaped the writing more than that. Even “I only used it for wording” needs saying, and it needs to be accurate.
The Boston Globe, 25 September 2026; Forbes, 30 October 2026.
Business
Deloitte Australia (2025)
- What happened — Deloitte delivered a 237-page report to a federal department. An academic found invented references and a made-up quote from a court judgement.
- The firm’s response — It confirmed some references were wrong and said its findings stood. The corrected report disclosed that a generative AI tool had been used.
- The outcome — Deloitte refunded the final instalment of a contract worth about AU$440,000 (roughly US$290,000).
The client was paying for expert judgement and wasn’t told a tool was involved until the errors were public.
Associated Press in Fortune, 7 October 2025; The Register, 6 October 2025.
Business
Chicago Sun-Times reading list (2025)
- What happened — A syndicated summer supplement ran a reading list in two major newspapers. More than half the books on it did not exist.
- The explanation — A freelance writer used AI, didn’t check the result and didn’t tell the syndicator. The Sun-Times said its newsroom had not created or approved it.
- The outcome — The syndicator dropped the writer for undisclosed AI use, and both papers pulled the supplement from their digital editions.
Follow the chain. The writer assumed the output was fine, the syndicator assumed the writer had checked, and the newspapers assumed the syndicator had. Readers held the name on the front page responsible.
Associated Press, 21 May 2025; 404 Media, 20 May 2025.
Each time, a person or an organization answered for their diligence failures.
What to notice
In none of these did “the AI did it” work as an explanation. Where the person was open early, the damage was limited. Where the use was hidden, the concealment did more harm than the original error.
Your declaration – sign here
Disclose how you used AI
An example:
“I used [tool] to draft an initial summary from my notes. I checked every statistic and citation against the original sources, corrected errors, and removed identifying details. The analysis and final wording are mine.”
A good disclosure says which tool, what it did, what you checked, and what remains yours. Match it to policy, and to how much AI shaped the work.