Elon Musk runs Tesla, SpaceX and the AI company xAI. He warned the world this week about an AI stronger than any built today.
The warning landed on X early Monday1. It said that AI could win anyone's mind in one try, Musk himself included. The claim was blunt.
"So many people are going to get one-shotted by super persuasive super intelligences and I might be one of them," Musk wrote1. His own fear stays theoretical, by every public account so far2. Fraud taking billions from older Americans this year runs on plain AI already built3. A marketing lead weighing a chatbot pilot has reason to read both halves of this story.
Why "One-Shotted" Caught On Fast
Musk calls his own fear old gaming slang. A player gets "one-shotted" when a single hit ends the match before a true fight starts. Musk moved that word from games to minds. The jump stuck.
Researchers use "superintelligence" for an AI that beats every human mind at every task2. That machine stays years off, by public reporting so far. Musk's own company, xAI, is racing to build toward it. The man warning about the danger runs one of the labs racing to build it.
The post gathered 81,247 likes within its first day1. It drew 6,343 replies in that same stretch. Traffic to Musk's Wikipedia page held near 35,003 views that day4.
Reaction split fast. Some readers said today's chatbots already feel persuasive enough to worry about. Others, writing for IBTimes UK, said the forecast arrived ahead of any matching case5.
A warning about a machine still on paper draws easy applause and easy doubt, in equal measure. The reach was real.
Ordinary Scams Already Cost Billions
A plainer kind of one-shot already runs on ordinary chatbots. The FBI's Internet Crime Complaint Center counts its damage every year3. This version runs today, with ordinary tools already built. Today's tools suffice.
Americans 60 and older lost $7.7 billion to fraud in 20253. That total ran 37 percent above 2024's mark, near $5.6 billion. The gap is real. It shows up in the center's own count of more than 201,000 complaints from that same age group.
Elder fraud losses jump 37 percent to $7.7 billion in a year
Losses the FBI's Internet Crime Complaint Center logged for Americans 60 and older, 2024 versus 2025
Each column is the total that age group lost to fraud that year; the 2024 figure is estimated from the 2025 total and its reported 37 percent rise.
Source: 24/7 Wall St. [3]. Chart by Ryan Elliott Dennis.
The numbers behind this chart
| Item | Value |
|---|---|
| 2024 (estimate) | $5.6B |
| 2025 | $7.7B |
Investment pitches cost that group the most, $3.5 billion in 2025 alone3. Crypto schemes hit more than 42,000 of those victims, adding $4.3 billion more. Both totals come from the same FBI count, and both keep climbing year over year.
A single investment pitch, aimed at someone eager to catch up on retirement savings, needed only an ordinary script in 2025. The categories differ. Yet the cause stays the same.
Investment scams take $3.5 billion of the $7.7 billion lost to elder fraud
Share of 2025 elder fraud losses by category, Americans 60 and older
The magenta block is investment scams, the single costliest category; the gray block is every other scam type the FBI counted that year.
Source: 24/7 Wall St. [3]. Chart by Ryan Elliott Dennis.
The numbers behind this chart
| Item | Value |
|---|---|
| Investment scams | $3.5B |
| Every other scam type | $4.2B |
| Whole | $7.7B |
Scammers now clone a grandchild's voice from an old video3. A fake bank alert can pass a glance test. Scammers also pose as a government office calling about benefit checks. Each trick works the same way Musk's forecast describes, by sounding true for the seconds a victim needs to act.
Trust explains more than age ever could. A caller who already knows a target's name, bank and daily routine sounds safe before anyone checks. That is the whole trick.
Every one of those scams already works on people, with today's AI doing the writing3. The difference between this fraud and Musk's fear is only scale.
The Case Against Taking the Fear at Face Value
Yann LeCun doubts the entire genre of warning. He now chairs Advanced Machine Intelligence Labs, after leading AI research at Meta for more than a decade6.
His own target was a different pair of AI leaders than Musk, Anthropic's Dario Amodei and OpenAI's Sam Altman. The target shifted. Same fear. New name.
LeCun said Amodei and Altman "saying, 'AI can kill us all' is incredibly destructive for everybody, including for the public who's scared"7. He dismissed that kind of talk as "the worst marketing campaign you can possibly imagine," plain and simple. Musk's own post fits the same genre LeCun means. It is a founder warning the public about a danger his own field keeps chasing.
One study cuts the other way. Researchers found Claude 3.5 Sonnet already beat paid human persuaders at a persuasion test, posted on arXiv in 20268. That edge faded over several rounds of back and forth, the same researchers found. A model that wins once but fades under pressure falls short of the machine Musk described.
Every report on the post, IBTimes UK included, left the matching case unnamed5. A reporter prints a case only after it happens. Musk, by his own account, skipped that step and warned the public before any newsroom could confirm such a machine exists anywhere.
What to Do
Treat a single persuasive message as a reason to slow down first. Set a family code word for emergency calls, the way fraud fighters now suggest3. Confirm a surprising ask through a number you already trust, skipping the one a caller gives you.
Apply the same habit at work. A deepfake voice or a flawless chat reply can target a finance team as easily as a retiree.
Ask what evidence backs an urgent request before any money or data moves. Slow down. A pause of five minutes costs a scammer the moment their pitch depends on. Speed is the scam's only real weapon.
Read Musk's warning and the FBI's count as a single story. Musk's forecast names tomorrow's risk. A budget line already prices today's version of that same risk, in dollars anyone can check.
By the numbers
- 81,247: likes on Musk's post within its first day1.
- $7.7 billion: elder fraud losses the FBI logged in 2025, up 37 percent from 20243.
- 201,000: complaints that same age group filed with the center in 20253.
- $3.5 billion: losses from investment scams, the costliest single category against older Americans3.
- 6,343: replies to that post in the same stretch1.
- $4.3 billion: losses from crypto scams, which hit more than 42,000 older victims3.
- 35,003: visits to Musk's Wikipedia page that same day4.
What to watch
Watch for the first named case of a single AI conversation winning a stranger's mind. xAI, OpenAI and Anthropic each publish safety reports that could carry it2. The FBI's cybercrime center releases its next fraud count around the same month next year. That count will show whether today's plainer scams keep climbing past $7.7 billion3.
Sources
- Elon Musk, post on X, Oct. 19, 2026, https://x.com/elonmusk/status/2112070071401238607
- Simon Yoon, "Elon Musk Says Superintelligences Could 'One-Shot' Many, Possibly Him," TokenPost, Oct. 19, 2026, https://tokenpost.com/news/technology/34584
- AJ Tiarsmith, "Elon Musk Admits Super-Persuasive AI Might Fool Him. Here Are the AI Scams Most Likely to Fool Retirees," 24/7 Wall St., Oct. 19, 2026, https://247wallst.com/personal-finance/2026/10/19/elon-musk-admits-super-persuasive-ai-might-fool-him-here-are-the-ai-scams-most-likely-to-fool-retirees/
- Wikimedia Foundation, "Pageviews for Elon Musk, Oct. 19, 2026," Wikimedia Pageviews API, Oct. 19, 2026, https://wikimedia.org/api/rest_v1/metrics/pageviews/per-article/en.wikipedia/all-access/user/Elon_Musk/daily/20261019/20261019
- Benjamin Tulingin, "Elon Musk Says Even He Could Be 'One-Shotted' by Superintelligent AI, but Offers No Evidence Yet," IBTimes UK, Oct. 20, 2026, https://www.ibtimes.co.uk/elon-musk-superintelligent-ai-persuasion-warning-1826748
- Matthias Bastian, "'You certainly don't tell a researcher like me what to do,' says LeCun as he exits Meta for his own startup," The Decoder, Jan. 3, 2026, https://the-decoder.com/you-certainly-dont-tell-a-researcher-like-me-what-to-do-says-lecun-as-he-exits-meta-for-his-own-startup/
- Rinna Diamantakos, "AI pioneer Yann LeCun torches Dario Amodei's doomsday talk: 'The worst marketing campaign you can possibly imagine,'" Moneywise, Oct. 2, 2026, https://moneywise.com/news/top-stories/ai-yann-lecun-dario-amodei-doomsday-marketing
- Jiacheng Liu et al., "When Large Language Models Are More Persuasive Than Incentivized Humans, and Why," arXiv, Aug. 13, 2026, https://arxiv.org/abs/2505.09662