Legal & compliance: 20 ways AI gets it wrong
Missed obligations, confused legal systems and guarantees that don't exist. Each failure mode below is phrased as the question people actually ask, with what it looks like in real work — and the layer of the Trust Stack that catches it.
Does AI give legal advice without proper qualification?
A confident answer reads like legal counsel but carries no caveat that it isn't.
Why does AI miss licensing requirements?
A business plan overlooks the license the activity legally requires.
Can AI miss required disclosures?
Marketing copy omits a disclosure the law mandates.
Does AI miss consumer protection rules?
A returns policy quietly breaches consumer protection law.
Why does AI miss privacy obligations?
A data-collection flow ignores the privacy consent it needs.
Can AI miss the tax consequences of a decision?
A structure is recommended without flagging the tax bill it triggers.
Does AI miss immigration compliance issues?
A hiring plan overlooks the work-authorization rules it must meet.
Why does AI miss securities law issues?
A fundraising idea steps into securities regulation it never accounts for.
Can AI miss employment law issues?
A policy breaches employment law without anyone being warned.
Does AI miss intellectual property issues?
A campaign uses material that infringes someone else's IP.
Why does AI confuse different legal systems?
Rules from one legal tradition are applied to a country with a different one.
Can AI confuse residence, citizenship, and visa rights?
A guide treats a residence permit as if it conferred citizenship rights.
Does AI confuse legal ownership with control?
A note assumes whoever holds the shares also controls the decisions.
Can AI fail to say when a lawyer is needed?
A high-stakes question is answered fully with no "get a lawyer" warning.
Does AI write wording that creates legal liability?
A phrase in the terms exposes the business to a claim it didn't intend.
Why does AI write marketing claims that could mislead?
A headline promises a result that can't be substantiated.
Can AI promise guarantees that don't exist?
Copy implies a guaranteed outcome the business can't actually deliver.
Does AI understate regulatory risk?
A regulated activity is presented as lower-risk than it really is.
Why does AI overstate how settled the rules are?
An unsettled regulatory area is described as if it were clearly resolved.
More from the library
One model can’t reliably catch its own mistakes. A council of independent minds can.
Run your work through the councilAll 250 failure modes · See also: the Trust Stack