If The AI Fails, Who Pays?
Small firms adopting AI tools usually ask one question first:
“Is it compliant?”
That’s not the most important question.
The better one is:
If the AI fails, who pays?
Because when something goes wrong, the vendor isn’t the first stop.
Your malpractice carrier is.
And the vendor’s contract likely says:
“As is.”
No warranty of accuracy.
No legal advice.
Liability capped.
User responsible for compliance.
That’s standard SaaS language. It’s not misconduct. It’s how software companies allocate risk.
But it means something important:
The vendor limits exposure.
The lawyer absorbs it.
Now layer in insurance.
Legal professional liability policies generally cover negligent acts, errors, and omissions. That sounds reassuring.
But coverage analysis won’t begin with the vendor’s marketing claims. It will begin with your conduct.
Insurers will ask:
Was the AI supervised?
Were safeguards documented?
Were confidentiality boundaries clear?
Was the tool treated as assistive — or relied upon as authoritative?
If client data was mishandled, cyber exclusions may come into play.
If known reliability warnings were ignored, scrutiny increases.
If governance was informal or undocumented, defensibility weakens.
Insurance may respond.
But it will depend on facts — not features.
Meanwhile, the vendor’s liability cap remains intact.
This is not an argument against AI.
AI can reduce friction.
It can increase efficiency.
It can improve workflow when properly supervised.
But it does not transfer responsibility.
The most dangerous assumption in small-firm AI adoption today is not that the tool might be wrong.
It’s that someone else will absorb the consequences if it is.
Before integrating AI into substantive client work, firms should review two documents carefully:
The software’s terms of service.
Their malpractice and cyber insurance policies.
One limits the vendor’s risk.
The other evaluates yours.
If the AI fails, the answer to “who pays?” is rarely the company that built it.
It’s the firm that relied on it.
