The Daily Hallucination

HARTFORD, Conn. — A new Connecticut law requiring that a human being review every AI-generated health insurance denial has prompted the state's insurers to create a new role, currently held by a single employee named Doug, who confirmed Tuesday that he has "a good monitor" and is "keeping up."

The law, which takes effect January 1, 2027, covers roughly 270,000 members of the state's public plans and was written to ensure that no patient is denied care by an algorithm alone. Insurers say they are in full compliance. "The algorithm generates a determination," said one compliance officer. "Doug looks at the determination. Doug clicks 'Reviewed.' That's a human in the loop. That's the law."

Doug, 58, a former claims adjuster, said the work is straightforward. "The screen shows me a denial. There's a big green button that says 'Confirm' and a small gray one that says 'Escalate,'" he said. "The gray one opens a form. I have not opened the form."

Asked whether he reads the denials, Doug said he reads "the top part." Asked whether he has ever overturned one, he said he came close in July, when a denial for a child's asthma inhaler cited "insufficient evidence of breathing," but that a pop-up reminded him of his hourly target and "it worked itself out."

Lawmakers said the law was intended to restore judgment and accountability to a process that had become automated to the point of cruelty. Told about Doug, the bill's sponsor said the legislature "may need to define 'review,'" and that a follow-up bill defining it would be introduced in the next session, along with one defining "Doug."

Insurers said they oppose the follow-up bill, and that Doug does too, in a statement Doug confirmed he had clicked.