BS BULLETIN:
- AI researchers are sounding increasingly dire warnings that future superintelligent systems could threaten humanity through cyberattacks, bioweapons, critical-infrastructure sabotage or military systems.
- Former OpenAI and Anthropic researcher Jacob Coxon quit Anthropic this week, warning that frontier labs are racing toward systems that could become extraordinarily difficult to control.
- Experts interviewed by the New York Post say billionaires’ bunkers wouldn’t offer much protection in the most extreme scenario: if a superintelligent AI truly wanted humanity gone.
Apparently, artificial intelligence isn’t content with writing your emails, making your vacation itinerary, and occasionally insisting that there are three R’s in a word containing two.
Some of the people building the most powerful AI systems now warn that future versions could conceivably kill us all. And they aren’t necessarily talking about an army of Terminators marching down Main Street.
A new New York Post examination lays out several scenarios AI researchers fear could unfold if humans eventually lose control of a superintelligent system — from crippled electrical grids and financial markets to engineered pathogens and autonomous weapons.
The warnings have gained new urgency following the resignation of Jacob Coxon, a researcher who worked at both OpenAI and Anthropic. Coxon announced this week that he was leaving Anthropic because of concerns over the race to develop increasingly powerful AI.
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
— Jacob Coxon (@hilbertspaess) September 9, 2026
“These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources,” Coxon warned.
His concern isn’t merely hypothetical mischief by today’s chatbots. It’s the possibility of future AI systems becoming capable of improving themselves, operating autonomously, and resisting human attempts to control them.
And Coxon isn’t alone.
Evan Hubinger, who leads alignment stress-testing research at Anthropic, said he believes there is a greater than 10% chance AI could wipe out humanity within the next decade. There is no scientific consensus supporting a precise probability, however, and other AI researchers consider extinction scenarios much less plausible.
One potential nightmare begins with infrastructure.
UC Berkeley computer scientist Stuart Russell told the Post that advanced AI systems could potentially attack electrical grids, transportation networks and financial markets. In a worst-case scenario described by experts, an AI capable of sophisticated hacking could disrupt electricity and communications, interfere with water systems, disable financial services and manipulate humans into carrying out physical tasks on its behalf.
Then there’s biology.
MIT professor Max Tegmark warned that increasingly capable AI could dramatically lower the difficulty of designing dangerous biological agents. Other experts similarly identify AI-assisted biological weapons as one of the more serious potential risks, although the ability of current systems to produce such catastrophic outcomes remains contested.
Military applications provide another concern. AI is increasingly being integrated into intelligence analysis, targeting, and autonomous weapons. Researchers worry that handing progressively more decision-making authority to machines could eventually leave humans unable to understand — or override — critical decisions quickly enough.
And what about the billionaires building elaborate bunkers? Don’t count on them. Russell called the idea that an individual could simply hide from a hostile superintelligence “almost childish.”
Anthony Aguirre of the Future of Life Institute put it even more succinctly: “If AI wants humans gone, we’re gone. Your bunker’s not gonna save you.”
There is considerable disagreement over whether anything remotely this catastrophic will happen.
AI researcher Gary Marcus told Business Insider that while cyberattacks, biological weapons and disinformation present serious risks, he sees no realistic near-term path by which AI actually exterminates every human being. Other researchers argue today’s systems remain far from the kind of broadly capable superintelligence envisioned in these scenarios.
So nobody knows whether artificial intelligence will cure diseases, revolutionize productivity, usher in an age of abundance — or eventually decide humanity is an unnecessary software dependency.
But when the people actually building the stuff start publicly debating whether it might kill everybody, it’s probably worth listening. On the bright side, if AI destroys civilization, at least nobody will have to remember another password.
DBS WIRE SOURCES:
- New York Post — Experts warn AI could kill us all — here’s how an apocalyptic scenario could play out
- Associated Press — Anthropic researcher resigns with warning about dangers of AI development
- Wall Street Journal — How would AI actually kill us all?
- CBS News — Humans are ‘close to being outsmarted’ by superintelligence, AI expert says
- Axios — Here’s how AI could kill us all, if the worst fears come true
