Science & Tech · AI · 5 hrs ago
Sam Bankman-Fried puts chance of AI destroying humanity at about 25%
Sam Bankman-Fried told 14-year-old podcaster Eli Goldfine that he puts the chance of AI wiping out humanity at about 25%.
The former FTX chief gave the estimate during phone calls from a federal prison in Lompoc, California, where he is serving a 25-year sentence for fraud.
The estimate is a version of “P(doom,” the term for the chance of an AI-driven catastrophe, though Bankman-Fried said the figure depends on how people define catastrophe.
He described possible outcomes ranging from peaceful cooperation between people and AI to AI destroying humanity.
Bankman-Fried was an early investor in Anthropic, but his shares were sold for $1.3 billion to repay FTX creditors, including people harmed by the fraud.
He remains bullish on Anthropic, and he has asked the US Supreme Court to consider his appeal of his conviction.
Sam Bankman-Fried estimated that artificial intelligence has about a 25% chance of destroying humanity.
He gave the estimate in a phone interview from federal prison with 14-year-old podcaster Eli Goldfine.
The roughly 90-minute interview took place over calls on Saturday and Sunday.
Bankman-Fried said estimates depend on what people count as a catastrophic event, giving a possible range of 20% to 60%.
He also described possible peaceful coexistence between humans and AI, as well as the risk that AI could wipe out humanity.
- Who
- Sam Bankman-Fried gave the estimate in an interview with 14-year-old podcaster Eli Goldfine.
- What
- He estimated that AI has about a 25% chance of destroying humanity.
- When
- The interview took place on Saturday and Sunday; the article was published on November 19, 2026.
- Where
- Bankman-Fried spoke by phone from federal prison in Lompoc, California.
- Why
- He said the estimate depends partly on how people define a catastrophic event and whether humans and AI coexist peacefully.
This story does not have two clearly opposing sides.
Bankman-Fried and Goldfine conducted the interview over a series of phone calls.
Business Insider published its report on Bankman-Fried’s estimate.
- Estimate
- About 25% chance of AI destroying humanity
- Interviewer
- Eli Goldfine, a 14-year-old podcaster
- Interview length
- Roughly 90 minutes
- Location
- Federal prison in Lompoc, California
- Alternative probability range
- 20% to 60% depending on how catastrophe is defined
What people are saying
I would say, like, less than 50%, more than 10%. You know, somewhere in the 25% range.
So depending on that, I think I could answer anywhere between, like, 20 and 60 percent, depending on those sort of gray areas.
Some chance of synergistic work, driven largely by AI, but with human goals in mind — that will lead to massive, massive technological expansion.