I asked AI should we take AI doomsters seriously . this was the reply by Chat GPTYes — but not necessarily in the way the loudest “AI doomers” present it.There are really two separate questions:Are serious AI risks worth taking seriously? Absolutely. Current research points to real risks from cyberattacks, weapons, misinformation, fraud, concentration of power and increasingly autonomous AI systems. A 2026 study of 272 AI specialists found substantial concern across a range of these risks. �MIT Sloan +1Should we believe that AI is inevitably going to wipe out humanity soon? That's much less established. There is no expert consensus on the probability or timing of an extinction scenario. Even researchers who take existential risk seriously disagree dramatically about it. �arXiv +1And there's an important distinction between “this could happen” and “this is going to happen.”The current evidence is actually rather interesting. Stanford's 2026 AI Index says AI capability is advancing extremely rapidly while safety, transparency and governance are struggling to keep pace. It also reports that documented AI incidents are increasing. � That's a good reason for sensible safeguards.Stanford HAI +1But some of the more dramatic claims — AI becomes conscious, secretly improves itself, takes over the internet and kills everyone — involve a chain of assumptions about future capabilities that we simply don't know will come true.So I'd put it this way:Don't dismiss the doomsters. Don't automatically believe them either.Take the risk seriously, rather than taking the prediction as fact.And there's another wrinkle: some of the people warning most loudly are themselves prominent figures in the AI industry, while others argue that catastrophic claims are exaggerated or could even serve particular commercial or regulatory interests. That disagreement is itself worth keeping in mind. �The GuardianIf you want, I can also explain **what the most credible “AI could actually go horribly wrong” scenario looks like