Right, But Overruled
So there I was...
Right, but overruled. If you've been following my blog, you know that I am a dog lover. Dogs have been a part of my life since I was a young kid, starting with an amazing dalmatian who accompanied me to the bus stop every morning and faithfully waited for me to come home each afternoon. And long before my long series of St. Bernards, I worked with some amazing German Shepherds, all Air Force military working dogs during my time in the service. I was recruited to be the medic for the Strategic Air Command Emergency Service Team, essentially a quick reaction force for incidents on base, so I got to know all of the handlers and their dogs as they rotated in and out of the team.
My regular job was running a crew in a very busy ER at an Air Force regional medical center, and everyone in law enforcement and fire knew me. So it was no surprise when our crash phone rang and the dispatcher asked for me by name to handle a really difficult situation. One of the dogs had passed away in his kennel from intestinal torsion, and the other dogs were going nuts. They asked if I would come get him. "Of course," I said.
Since I was going to be leaving the ER temporarily short staffed during a night shift, I let the charge nurse know that I needed someone to cover. When I told her why she lost her mind, there is really no other way to put it. She told me in no uncertain terms I could not go. To make a long story short, I told her to write me up and I went to get the dog with one of the ambulances and brought him back to the hospital morgue. When our 12-hour shift ended at 0700, we headed to the commander's office to make our regular report. What followed was epic.
She laid into me the moment we walked through the commander's door, so I simply stood at attention and let her dig a very deep hole. After a couple minutes of ranting, the Colonel raised his hand to silence her. "Major," he said quietly, "Sgt. Elmore is right, you are wrong. That animal is a canine service member with a rank. We treat it with all the courtesy and respect of a human service member. What Elmore did was show respect to the dog, its handler, all of law enforcement, and everyone else on this base that depends on this hospital to act honorably. Apologize."
As an NCO, defying the instructions of an officer is a great way to find yourself devoid of stripes, but I understood the rules, what was at stake, and my duty in both the ER and on the EST. I made it clear to the Major that I would face the music for my decision, and I did. I also graciously accepted her apology and made an effort to remove any friction between us afterwards. It agonizes me to see decisions to do the right thing get overruled by someone with explicit authority and none of the moral authority. And that is at the heart of an AI lawsuit filed yesterday in a federal court in San Francisco.
Yesterday the province of British Columbia sued OpenAI and its chief executive, Sam Altman, in federal court in San Francisco. The suit stems from a shooting this past February in the small town of Tumbler Ridge, British Columbia, where eight people were killed, five of them children between twelve and thirteen years old, along with an education assistant. More than two dozen others were wounded. I am not going to go deeply into the background of what happened that day, because the impacted families have enough to deal with. What matters for this story is a decision that was made months earlier, far from that school, in a Bay Area office.
According to the province's filing, OpenAI's own systems flagged the shooter's account in June of 2025, marking it as belonging to a user who might be planning to commit acts of violence. The company's human reviewers looked at it and did exactly what you would want a competent person in that seat to do. They recommended referring the account to the Royal Canadian Mounted Police. Someone inside OpenAI saw the danger clearly and said the right thing out loud: call the police. And then, the lawsuit alleges, leadership rejected that recommendation, quietly deactivated the account, and notified no one. This past April, Altman himself said he was "deeply sorry that we did not alert law enforcement to the account that was banned in June." The province reduces it to a single unbearable line. "One telephone call" could have prevented the tragedy.
Yesterday I wrote about the competent chair, the trained person who says hold and is listened to, and a war that did not start because someone paid attention. This is the very same situation and lesson with a tragic ending. The chair was filled. The right person was in it. They reached the right conclusion. And someone with more rank reached down and overruled them. I have been on the wrong end of that exact kind of overrule, and I just recounted one night when it happened to me.
The difference between my night and theirs is the Colonel. When my Major forbade me from doing the decent thing and threatened me over it, a man with both authority and conscience sat in his chair and set it right. The hospital commander had my back. The reviewer inside OpenAI who said to call the Mounties had the opposite. The override stood. Nobody above it reversed a wrong call before it was too late.
The charge nurse on duty with me that night had every bit of the authority to stop me and not one ounce of the moral authority to be right, and that gap is the whole subject of this blog. A chair with rank in it is not the same thing as a chair with judgment in it, and when the two collide, everything comes down to whether the person with the oak leaf will defer to the person with the understanding. Most of what I have written here has been about chairs left empty. This is about a chair that was filled by precisely the right person, overruled by someone who had the power to do it and no standing to be right, and then left with no one to rescue the decision.
It is worth saying that the province is suing Altman by name, not only the company. That is consistent with everything I have argued in this chair about who owns the consequences of an AI. The person behind the machine does not get to stand behind it. And the most damning thing in this whole matter is not an allegation at all. It is an apology. Altman has already conceded that the call was never made.
Let me be careful, because these families are owed accuracy more than they are owed our anger. All of this is alleged in a complaint that OpenAI has not yet answered. Whether one phone call would truly have stopped what happened is something no one can prove, and a defense team will spend a great deal of effort arguing exactly that point. It is a hard and genuinely novel case, a Canadian province suing an American company in an American court over a duty the law has barely begun to understand let alone define. But you do not need the verdict to see where this is going. A safeguard existed. A human being read the warning correctly. The safeguard did its job right up until a person with authority tragically said no.
Every company racing to put AI systems into the world is building chairs like that reviewer's, seats where a person will one day look at a warning on a screen and know that the right thing to do is pick up the phone. The machine did not make the decision that followed the warning in June. A person did, and the person above that reviewer overruled them. Whether the next warning ends differently depends entirely on whether the people with the authority will honor the people with the experience and judgment, or overrule them for reasons that play out very differently in a courtroom than in a team meeting. Eight people in Tumbler Ridge are owed an answer. I had my Colonel. They never got theirs.
In the US, call or text 988 for the Suicide and Crisis Lifeline. In Canada, call or text 9-8-8 for the Suicide Crisis Helpline. Both are free, confidential, and available 24 hours a day.
#AI #Accountability #MoralAuthority #Safety #OpenAI #HumanInTheLoop #TheEmptyChair #STIW