Moonshots with Peter Diamandis · Sep 27, 2026 · recorded Sep 25, 2026 · 51m · 38m saved
With Emad Mostaque, Founder of Intelligent Internet and former Founder and CEO of Stability AI
Bernie Sanders wants a permanent ban on artificial superintelligence, backed by a corporate death penalty and 20 years in federal prison for individuals.
Most of the argument over that proposal is about whether AI is dangerous. Emad Mostaque made a narrower point: the US already tried to ban encryption decades ago, and lost, because the courts treated code as speech.
"ASI, superintelligence, it's mathematics, and it's unconstitutional to ban it because it's free speech."
Emad Mostaque, who founded Stability AI before starting Intelligent Internet, joined the four regular panelists on Moonshots with Peter Diamandis at the show's first live event. The panel covered Jensen Huang's attack on AI doomers, the Xi-Trump AI talks, the rebranding of AGI as superintelligence, a run of 12 model releases in 22 days, and Anthropic's new wet lab.
The full episode is covered here so you can skip it. 51 minutes of audio, 13 minutes of reading.
Here are the 9 takeaways that matter.
Key Takeaways
Several separate AI risks are being merged into one fear of AI breaking out and killing everyone, when existing rules could handle most of them
AI that runs courts, hospitals and roads should be treated as a utility, with a liability waiver the frontier labs should not get
Most useful AI will be "execution" AI, not the frontier models that invent new physics
Banning superintelligence would be unconstitutional, Mostaque said, because it is math, like encryption
Sanders' plan carries a corporate death penalty and 20 years in prison
Claude Opus 5.5 is the first truly competent model, in Mostaque's view, after an Opus 5 he called "completely unhinged"
A general-purpose model drove a Toyota Corolla through a 130-meter cone course on its second try
Anthropic reported that 950 Claude agents found an unknown enzyme in 21 hours, days after its CEO called for a slowdown
Biology's hardest problems should get 1,000 agents each, with everything open sourced
1. The Doomer Backlash
Peter Diamandis, founder of XPRIZE and Singularity University, opened on what he called a pandemic of fear. Two weeks earlier, Anthropic's Dario Amodei had published a blog post on AI risk, and Sam Altman, Elon Musk and Demis Hassabis had agreed with him. Two nights before the recording, Altman and Amodei told the Security Council that the world is in the midst of an existential threat. Diamandis said the danger he sees is humans using AI with bad intent, not AI going rogue on its own.
He played two clips. Nvidia CEO Jensen Huang said nobody is building more compute than the people asking to be slowed down, and called Geoffrey Hinton's 10% estimate of societal destruction ungrounded in science or research. Meta CEO Mark Zuckerberg said he did not think the labs need industry-wide coordination, and that each lab should slow down internally when it sees a problem.
Mostaque said the real near-term risk is people misusing AI
He said many recent OpenAI incidents came down to poor security rather than a model breaking out:
"But a lot of the OpenAI outbreaks and things were bad security protocols. It was bad infrastructure on their part." — Emad Mostaque
"So I think we've got a series of dangers coming and they're all being conflated into one thing, which is AI breaking out and suddenly killing us all, without actually looking at the individual things and putting in place existing regulations and enforcing them." — Emad Mostaque
The other panelists went further:
Alexander Wissner-Gross, a computer scientist and founder of Reified, said a moral panic over nuclear energy after World War II cost at least 50 years of progress, and a panic over superintelligence could cost another 50. He said the worst outcome would be deceleration
Salim Ismail, founder of OpenExO, said the odds of a good outcome outweigh doom by about a thousand times. He said people have augmented themselves with technology for a long time, from glasses to childhood vaccines
Dave Blundin, founder of Link Ventures, said misuse is a real threat with real solutions, but getting stuck on doomerism is completely dysfunctional
