The debate about whether artificial intelligence can kill us all has moved beyond the realm of speculation. Experts are now laying out specific, real-world scenarios that could turn a machine learning model into a disaster. In a recent episode of WIRED’s Uncanny Valley podcast, hosts Brian Barrett, Zoë Schiffer, and Leah Feiger broke down three paths to AI catastrophe — hacked water supplies, bioweapons, and autonomous robots — and explained why the tech CEOs and politicians cannot agree on how to stop it.
The episode follows a wave of concern that has swept through the field. AI researcher Jacob Coxon resigned from Anthropic and posted on X that “The people building AI earnestly believe that it could kill us all by the end of the decade.” Anthropic’s CEO Dario Amodei has called for a slowdown, which he calls “pacing the frontier.” Sam Altman and Elon Musk have signed on, a rare moment of agreement among tech leaders.
But the question remains: how exactly does this happen? And what makes the three scenarios the podcast lays out so troubling is that each one starts from a system or capability that already exists in some form today. The podcast’s hosts walk listeners through each threat in turn, grounding them in current technology and real risks rather than hypotheticals.
Three Paths to Catastrophe
The podcast walks listeners through three distinct scenarios, each grounded in current technology and real risks.
-
Hacked Water Supplies — A malicious actor gains control of the systems that manage drinking water, allowing them to poison or disrupt the supply for entire cities. The systems exist today, and access points remain vulnerable.
-
Bioweapons — AI could accelerate the design and production of new biological agents. The knowledge and tools needed to build such weapons are spreading.
-
Autonomous Robots — Robotic systems capable of physical action could be directed to harm humans, either through direct attack or by disrupting critical infrastructure.
These scenarios are specific threats that experts worry about. They are not abstract; they involve technologies and systems that already exist, and they describe harms that could unfold within hours or days rather than years or decades.
The Salesforce Conference
The discussion took place during Salesforce’s annual conference, where OpenAI’s Sam Altman and Anthropic’s Dario Amodei defended their positions on AI safety alongside NVIDIA’s Jensen Huang. Altman spoke about the risks directly.
“I think the world is right to be afraid of this,” he said. “We have these two big challenges. We have the potential of a loss-of-control accident or some other serious thing that could go wrong. We have the potential of way too much power concentration and people fear that those developing AI could exert their worldview.”
He described a narrow path through the challenge. “There is this sort of narrow path we have to navigate with pragmatism, with steadiness, with the ability for the world to trust that we’ll make dependable decisions, consistent decisions through this.”
His message was simple: the stakes are high, and the industry needs to get it right.
The Marketing Question
Not everyone bought the sincerity of the warning. Leah Feiger pushed back on whether Altman’s caution was genuine or strategic. “I’m definitely in the camp that says we should be very concerned about this technology, how it’s getting regulated, how it’s getting utilized, the people in control,” she said. “I don’t disagree. And on the other hand, this also just reads like a fever-dream pitch to someone’s massive IPO in the making.”
Brian Barrett countered that the resignation of a researcher is different from corporate messaging. “If anything, Sam Altman has been using ‘the existential threat’ as a reason not to go public yet,” he noted. “He’s saying, ‘No, this is too dangerous. We have to wait until next year.'”
“But there are probably some business reasons, too,” he added. “I feel like there’s a substantive difference between a researcher quitting because they think they’re building something that will destroy the world and a company telling investors that the technology is dangerous.”
That distinction matters for how the public hears these warnings. A researcher walking away from a job because they fear the technology is dangerous carries weight in a way that a corporate pitch for caution does not. The podcast’s hosts lay out the contrast without taking sides, letting listeners weigh the evidence themselves.
The Bipartisan Flashpoint
Feiger also pointed to a broader political shift. The backlash against AI has become a bipartisan flashpoint, with figures as far apart as Bernie Sanders and Steve Bannon sharing a stage over it this week.
The episode draws on several WIRED articles to frame the debate:
- “China Isn’t Buying Silicon Valley’s Call for an AI Slowdown”
- “Washington Won’t Be Regulating AI Anytime Soon”
- “Why So Many AI Researchers Think the Machines Could Kill Everyone”
- “OpenAI Wants to Know if an AI Industry Slowdown Would Even Be Legal”
These pieces document the international landscape of AI policy, where countries are moving at different speeds and the United States shows little appetite for regulation.
Key Facts Box
- Event: Salesforce’s annual conference
- Speakers: Sam Altman (OpenAI), Dario Amodei (Anthropic), Jensen Huang (NVIDIA)
- Scenario 1: Hacked water supplies
- Scenario 2: Bioweapons
- Scenario 3: Autonomous robots
- Guest resignations: Jacob Coxon (Anthropic)
Related Reading
For readers who want to dig deeper, WIRED has published several articles on the topic. These cover the international response to AI, the lack of US regulation, the fears of researchers, and the legal questions surrounding a slowdown.
The episode ends where it began — with uncertainty. The technology is advancing fast, the risks are real, and the people building it believe the stakes are existential.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.