An Anthropic researcher recently stated that artificial intelligence has more than a 10% chance of killing all humans. The statement landed with the same fanfare as a software engineer mentioning they found a mild memory leak in the break room.
This is remarkable not because the claim itself is new — AI researchers have been casually dropping extinction scenarios into their LinkedIn posts and conference presentations for years now — but because we have collectively decided that a one-in-ten shot at human annihilation is workplace small talk.
Imagine your structural engineer mentioning over coffee that the bridge you drive on daily has a 10% failure rate that could kill everyone on it. You would probably not nod and move on to discuss the weather. Yet when the same probability gets attached to the technology reshaping every industry, the response is a gentle nod and a “fascinating research direction.”
The real absurdity is not that AI researchers worry about catastrophic outcomes. It is that we have normalized discussing those outcomes as if they were quarterly performance metrics. Safety warnings have become so routine that a 10% extinction probability barely cracks the news cycle. It sits somewhere between a new model checkpoint and a blog post about prompt engineering.
This is what happens when the stakes become too large to process emotionally. We convert existential dread into data points, then file it away next to the other thousand urgent things demanding attention. The researchers are not being reckless — they are being human. We just happen to be the species they are worried about deleting.