The New Liability Defense: "The AI Told Me To"

Reason's piece pushes back on a growing habit: patients and even practitioners deflecting responsibility for bad health outcomes onto AI tools, rather than owning the decisions made with them. The underlying detail set here is thin -- a single opinion piece -- but the framing itself is the story worth tracking.

This is the opening move of a pattern that shows up every time a new decision-support tool enters a high-stakes field: first the tool gets blamed for outcomes it merely informed, then regulators get pressured to treat the tool as the responsible party, then liability law scrambles to catch up. It happened with GPS and driving negligence, with algorithmic trading and flash-crash blame, with autopilot and aviation incidents. The pattern always ends the same way -- courts and institutions eventually re-anchor responsibility on the human who acted on the tool's output, but not before a messy interim period where "the algorithm did it" becomes a convenient shield for bad judgment, and a convenient cudgel for people who want the tool regulated out of existence.

The SAL read: if your business is integrating AI into any decision chain -- medical, financial, operational -- document that a human reviewed and owned the call, because "the AI told me" will not hold up as a liability defense any better than "the GPS told me" did.

For operators, the practical exposure isn't the AI itself, it's the absence of a paper trail showing human judgment was applied on top of it -- fix that now, before a bad outcome forces the question in front of a lawyer or a regulator.