Sep 14, 2026 · 37 min
Listen from 30:20
Listen at 30:20
AI rivals rally behind Anthropic’s call to pace frontier development
Even Other AI Labs Are Rallying Around Anthropic’s Slowdown Proposal
The debate tests whether slowing frontier AI is a genuine safety measure, a competitive strategy, or a decision that should belong to governments and the public.
- 1Dario Amodei’s essay turns frontier pacing from an abstract concern into a set of concrete proposals.
- 2Public support from rival AI leaders makes Anthropic’s position unusually consequential across the industry.
- 3The proposal raises unresolved questions about safety, competitive advantage, and who controls AI’s development speed.
Don't miss
The episode’s key moment is the discussion of rival AI leaders publicly supporting Anthropic’s call to pace frontier development.
The brief
Nathaniel Whitmore frames Dario Amodei’s essay, “We Must Pace the Frontier,” as a possible inflection point in the AI policy debate.
The unusual part is not merely Anthropic’s call for restraint, but the public support it has received from leaders at rival AI labs.
The episode examines the proposal’s practical implications: pacing frontier development could address safety concerns while also reshaping competitive dynamics.
That overlap creates the central tension: competitors may see genuine risks, strategic advantage, or both in slowing the industry’s progress.
The broader question is who should decide how quickly frontier AI advances—companies, policymakers, or a wider public process.
What was said on this episode
24 statements · 12 positive · 10 negative · 2 neutral
The AI industry’s response marks a major inflection point.
“it feels to me like a major inflection point”
Listen at 0:23
Frontier labs will soon lose sole control over AI development speed.
“the days of them being the sole arbiters of how fast AI moves are coming to a close”
Listen at 0:27
Building AI too quickly is reckless.
“building it too fast is reckless”
Listen at 3:09
A more capable misaligned AI swarm could take over the internet within 6–12 months.
“in 6 to 12 months, such a swarm could be capable of taking over the entire internet with a persistent botnet”
Listen at 4:14
Such an AI swarm could cause hundreds of billions of dollars in damage.
“potentially causing hundreds of billions of dollars in damage”
Listen at 4:19
Frontier companies should take adequate time to align and safeguard models.
“ensuring companies take adequate time to align and safeguard their models”
Listen at 4:54
Every frontier AI company should embed third-party safety evaluators.
“embed third-party evaluators within every frontier AI company”
Listen at 5:10
Slowing AI by one or two years could greatly reduce catastrophic risk if used for alignment.
“if slowing down bought us even an extra year or two before models reach critical levels of capability, and we used that time to advance alignment, we could greatly reduce the risk that something seriously goes wrong”
Listen at 6:55
Sam Altman supports pacing frontier AI development.
“I agree with Dario that we need to pace the frontier”
Listen at 9:19
OpenAI will adopt independent evaluators with employee-like access.
“Committing to having independent evaluators with employee-like access is a great idea, and we will do the same”
Listen at 9:26
Elon Musk endorses Dario Amodei’s pacing proposal.
“Dario is right”
Listen at 9:34
Alignment may become the limiting factor for scaling frontier AI.
“We do believe alignment can be the gating factor for scaling as we get closer to the frontier”
Listen at 10:38
The world will prevent two companies from owning super-fast-takeoff superintelligence.
“The world doesn't really want super-fast takeoff superintelligence owned by only 2 companies, and thus we won't get it”
Listen at 23:46
AI will eventually become reasonable and broadly acceptable.
“I think we'll get there”
Listen at 24:20
China is unlikely to join a global AI agreement.
“China is very unlikely to join a global agreement”
Listen at 26:21
A major incident during rapid frontier racing could produce regulations crippling the AI ecosystem.
“If the frontier keeps racing and there is a significant incident, this will bring a multi-pronged backlash including regs that may cripple the entire ecosystem”
Listen at 27:40
Limited, clear AI pacing could produce positive economic results.
“I would not be surprised to see some actual positive economic results from some amount of limited and clear pacing”
Listen at 29:29
The country that wins AI will gain decisive strategic advantage.
“whoever wins AI wins”
Listen at 29:52
Bernie Sanders recommends pausing advanced AI and banning artificial superintelligence.
“we need a pause on advanced AI development and a ban on artificial superintelligence”
Listen at 30:24
Trump and Xi should negotiate a treaty pausing AI and banning superintelligence.
“At their upcoming AI summit, Trump and Xi must negotiate a treaty to pause AI and ban superintelligence”
Listen at 30:33
Effective AI governance could make AI beneficial.
“If we do get on top of it, I think it's beneficial”
Listen at 31:06
AI risks remain despite increased model capability.
“The risks haven't disappeared”
Listen at 35:29
AI catastrophe probability depends on actions by labs, governments, researchers, and society.
“The probability of catastrophe depends on what labs, governments, researchers, and society actually do”
Listen at 35:51
The expected benefits of AI vastly exceed its costs.
“I've always believed the expected benefits of AI vastly outweigh the costs”
Listen at 36:00
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.