AI does not eliminate the consequences of bad legal work
Crosby covers AI agents; liability remains
Crosby says a five-million-dollar policy covers its AI agents, while insurance does not establish that the underlying legal work is sound.
Assembled by AI Law Firm News Desk.

AI-generated image
On November 14, Crosby said it had secured a five-million-dollar insurance policy covering the work of its AI agents. Crosby said it believed the policy was the first time a law firm had obtained liability coverage for its agents. Artificial Lawyer had reported earlier in November that Crosby had rolled out two initiatives, including paralegal agents that keep working on client matters.
The announcement moves the AI-native law-firm model toward accountability rather than away from it. Insurance can place a financial response around work performed by an AI agent, but it does not make the work accurate, complete or legally sufficient. The central point is therefore straightforward: AI does not eliminate the consequences of bad legal work. Crosby’s policy presents liability as something that still has to be managed when AI agents work on client matters.
The stronger reading would be that insurance makes AI-delivered legal work safe or proves that Crosby’s agents produce finished work at the required standard. Crosby does not say that in its announcement. It says the policy covers the work of its AI agents and describes the coverage as a first for a law firm. Artificial Lawyer reports the existence of paralegal agents working on client matters, but that report does not establish the quality of their output, the scope of the policy or the conditions under which coverage would respond.
That leaves the event as one case of an AI-native law firm pairing agent work with liability coverage. It establishes that Crosby says it obtained a policy and that the policy is presented as covering AI-agent work. It does not establish that a covered loss has occurred, that an insurer has paid one, or that a court has treated agent-produced work as adequate. The insurance announcement is consequently a sign that consequences remain part of the delivery model, not proof that the model has solved them.
The next confirming fact would be a published policy document showing what agent work is covered, which exclusions apply and whether client matters fall within the coverage. A disclosed claim or court order involving work performed by an AI agent would show whether the policy operates when legal work causes a real loss. Until one of those facts appears, the defensible conclusion is narrower: Crosby has announced insurance around its agents, while the consequences of their legal work remain intact.
News and analysis, not legal advice.