Technology
Humans Are Responsible. Not the Tool.
British Columbia wants OpenAI to answer for the Tumbler Ridge massacre. The Antigloss has a simpler question: what happened to the humans whose actual job was to recognize danger and intervene?
British Columbia is suing OpenAI and Sam Altman over the February massacre at Tumbler Ridge Secondary School. The province alleges that OpenAI detected disturbing conversations involving Jesse Van Rootselaar, disabled one account, but failed to notify police before she later carried out the attack. OpenAI disputes the implication that it ignored a clearly reportable threat.
That lawsuit begins remarkably late in the story.
Van Rootselaar was already known to police. Officers had attended the family residence repeatedly over concerns about her mental health. She had been apprehended under British Columbia’s Mental Health Act on multiple occasions and taken to hospital for assessment at least twice. Police later acknowledged that they did not know whether she was receiving treatment when the massacre occurred.
Firearms had also previously been seized from the family residence and later returned after an appeal, although RCMP subsequently established that the principal weapon believed to have been used in the school attack had never itself been seized.
Those facts point toward the people and institutions with actual real-world powers: family, police, doctors and the mental-health system.
They could observe behaviour outside a text box. Police could attend the home. Doctors could assess her. British Columbia already had legislation allowing involuntary psychiatric intervention under defined circumstances. Family members lived with the consequences of her behaviour rather than reading fragments of it through software.
That is where responsibility should begin.
Instead, the newest target is the information tool.
A Library With Better Search
Suppose Van Rootselaar had never opened ChatGPT.
She could have used Google.
Before Google, she could have used a library.
Information about firearms, criminal cases, police procedure and previous mass shootings did not materialize when generative AI was invented.
A novelist can research exactly the same subjects. So can a journalist, criminologist, screenwriter or historian.
A writer may ask how a shotgun functions because a fictional character needs to use one convincingly. The same writer may research police response to a school shooting because that is the plot.
Another person may ask identical questions with criminal intent.
The information does not reveal which person is sitting at the keyboard.
That distinction exists inside the human being.
AI can estimate intent from patterns. It cannot read it.
Follow the Logic Backward
If using ChatGPT makes OpenAI responsible, where does the chain stop?
At the smartphone?
The computer manufacturer?
The operating-system company?
The internet provider?
The processor manufacturer?
The company that made the weapon?
Every one of those products can become part of the physical chain leading to a crime.
We normally understand why that is not enough.
A tool being used is not the same as a tool deciding.
The shotgun does not develop murderous intent because somebody pulls its trigger. A library does not become an accomplice because a murderer once borrowed a book. A search engine does not acquire moral agency because somebody typed a disturbing query.
Artificial intelligence should not become the magical exception simply because it is newer.
The Alternative Is Surveillance
There is also a price to British Columbia’s preferred direction.
If AI companies face enormous liability whenever a disturbed user later commits violence, their safest corporate response is obvious: lower the reporting threshold.
Flag more conversations.
Inspect more context.
Block more accounts.
Send more questionable cases to police.
That sounds comforting until legitimate research starts looking suspicious.
The novelist writing a massacre scene gets flagged.
The journalist researching extremists gets flagged.
The historian examining terrorism gets flagged.
Police then receive more reports and must determine which describe genuine preparation and which describe Chapter Twelve.
That does not eliminate uncertainty.
It moves uncertainty into a surveillance system.
And the same public demanding that technology companies identify suspicious speech will eventually ask why private conversations are being monitored and handed to law enforcement.
You cannot simultaneously demand maximum monitoring and maximum privacy.
Start With the Humans
Eight innocent people were killed in Tumbler Ridge. The investigation remains active, including analysis of electronic devices, online accounts, witness statements and firearms evidence.
Every genuine failure should be examined.
But the chronology matters.
This was not a person whom society had never encountered until an algorithm noticed something strange.
Police had encountered her.
The mental-health system had encountered her.
Her family knew her in the real world.
Then, later, she used an AI service.
British Columbia now wants the newest technology in that sequence to shoulder part of the blame.
The Antigloss position is simpler:
**People are responsible for people. Institutions responsible for public safety should answer for how they exercised that responsibility. Information tools should remain tools.**
Yesterday the information came from a library.
Then Google.
Today ChatGPT.
**The interface changed. Human responsibility did not.**