The people building AI say it could kill us all by the end of the decade. That line has set off alarms across the internet. It has also drawn a sharp pushback from researchers who say the doom talk is a distraction from harms that are already here.
Subbarao Kambhampati, a computer science professor at Arizona State University, told Bloomberg that claims of out-of-control AI should not be used to sidestep accountability. "You start a process that goes haywire and causes damage," he said. "Even when your dog causes it, you are responsible."
That line landed. Online, the room split in a familiar way. Some people treated the "kill us all" warning as a serious red flag about a technology moving too fast. Others rolled their eyes and said it sounded like science fiction used to dodge real questions about bias, scams, deepfakes, and layoffs tied to AI systems already in use. A third group made jokes. The temperature was high, but the center held on one point: responsibility does not vanish because a tool is called AI.
The debate flared after an outgoing Anthropic employee posted on X that the people building AI earnestly believe it could kill us all by the end of the decade. The post went viral and fed a fresh wave of fear. Several experts pushed back, saying such talk lets companies frame safety failures as proof that containment is impossible, which shifts blame away from how experiments were run and who approved them.
Bloomberg reported that other experts argue this rhetoric allows companies to sidestep accountability for real-world impacts of their technology. Talking about AI as an autonomous entity reframes safety failures as evidence that containment is not possible. That framing reduces the focus on how tech companies may have conducted their experiments unsafely, according to Kambhampati. He said claims of out-of-control AI should not be used as a way to dodge responsibility for damage.
The documented harms are not abstract. Researchers and advocates have pointed to AI tools used to create non-consensual nude deepfakes of women, to misinform people, and to amplify scams. Others have flagged algorithmic bias and discrimination, data misuse, and AI-driven layoffs as material stakes being drowned out by existential fear. Timnit Gebru, founder of the Distributed AI Research Institute, told WIRED that the "machine-god narrative" behind AI extinction talk is meant to distract us from present harms, including autonomous weapons used in warfare and climate damage worsened by data-center demand.
This is where the argument gets sticky. If the public and regulators fixate on a future superintelligence that might turn on us, they may underinvest in rules and enforcement for the systems causing trouble today. Some critics say the doom framing invites regulatory capture around a narrow set of fantastical scenarios, while ordinary harms keep rolling. Others counter that long-term risk deserves serious attention and that slowing development could prevent catastrophe. Both sides can be true at once, and that tension is the whole problem.
Kambhampati's dog line cuts through the fog. You start a process that goes haywire and causes damage to my property, you are responsible, period. It is true even when your dog does it. That stance does not deny that AI could become more powerful. It insists that power does not erase liability. If a model discriminates, invents fake legal cases, helps someone build a dangerous weapon, or spreads harmful lies, the question should not be whether the model went rogue. It should be who built it so flimsy, who rushed it to market, who failed to test it, and who keeps using it after the harms become visible.
The internet knows this dance. A viral warning sparks panic. Experts call for focus on current harms. Companies promise safer testing methods. The cycle repeats. What changes is the temperature of the room. This time, the mood leaned toward quiet concern. The claim that AI risk narratives distract from accountability for real harms felt plausible and timely. It matched what many people have watched unfold: grand speeches about existential safety alongside product launches that ignore basic guardrails.
There is a path forward that does not require choosing between present and future. Liability and accountability can scale with demonstrated risks in real-world contexts, while independent scrutiny and high-quality evidence guide any limits on long-term scenarios. That means treating AI systems like other powerful technologies: subject to civil and criminal laws against conduct that threatens life and property, and to consumer-safety rules when products pose a danger. It also means more transparency about how models are built and tested, so outsiders can assess claims instead of taking marketing at face value.
None of this solves the hard questions about how fast to move or who gets to decide. It does restore a simple baseline. If your process causes damage, you are on the hook. The dog analogy is not cute. It is a reminder that agency belongs to the builder and the deployer, not the tool. The internet is tired of being told to fear the machine instead of the people who control it. That fatigue is showing in the replies. It is also showing in the quiet shift toward asking, again, who is being harmed today.
Explore: News