A consequence: AI governance gaps cost accountability

Legal Futures reported that a third of law firms are already using AI tools without a formal policy governing how AI is used, what data goes into it or who is accountable when something goes wrong. The report, published on October fourth, identifies an operational gap rather than a forecast: AI is already entering legal work while responsibility for its use remains unsettled.

That gap matters because AI does not eliminate the consequences of bad legal work. The Solicitors Regulation Authority says it has identified 42 reports involving possible misuse of AI, particularly inaccurate information and client confidentiality. The point is not that every AI-assisted task fails. It is that a finished legal service still needs a person and a process capable of checking accuracy, protecting client information and assigning responsibility when the process breaks down. Speed and scale do not answer those questions.

The same distinction appears in the treatment of privilege and supervision. Law360 Legal Industry reports that a New York City Bar Association ethics opinion cautions attorneys against recording nonclient conversations with AI tools and treats AI meeting recaps as a distinct business-record category. Reuters Legal reports that a federal judge warned increasing use of AI could create risks for clients and hinder junior-lawyer training, while declining to sanction attorneys over AI-generated errors in a filing. These reports support a narrower conclusion than AI makes legal work unviable: an AI-native model must build verification, confidentiality handling and accountable review into the service. The Legal Futures report does not establish that firms without formal policies have caused a particular client loss, that AI caused the reported matters, or that a leaner organization produces better economics. A stronger claim would require firm-level results showing how controls change error rates, privilege outcomes, client losses or margins. That is not shown here.

The next confirming fact is concrete: a firm-level policy that names permitted AI uses, controls the data submitted to the tool, assigns a responsible reviewer and reports the resulting error or confidentiality rate. A regulator finding that links a specific AI use to inaccurate information or a client-confidentiality failure would cut against the idea that deployment without governance can remain merely an internal efficiency choice. A court order or ethics opinion addressing AI-generated work would add a separate accountability test. Until those facts appear, the report establishes the cost of an unmanaged operating model, not the success or failure of AI-native legal delivery. The position holds at the level that matters: responsibility remains attached to the legal service even when software performs part of the work.