Palisade interviewed 22 current and former employees from OpenAI, DeepMind, and Anthropic about their personal views and fears around AI development. Today, we’re releasing the first batch of those interviews. Please watch and share.

Much of the world is trying to figure out what to make of Jacob Coxon's warnings. Are they marketing? Hype? A psy-op? Most people have no idea how seriously people at the labs take RSI or the possibility of human extinction. Full interviews at

We wanted to get AI lab employees on camera, to create a direct communication channel between the people working on AI, and the broader world. This is what they had to say:

has been a researcher at Google Brain, OpenAI, Google DeepMind, and UK AISI. He personally worked with Sam, Dario, and Demis. He thinks that the chance of human extinction is about a coin flip.

is widely regarded as one of the top two experts on mechanistic interpretability in the world. “Speaking as an interpretability expert, please do not rely on us to save you on the current trajectory."

Juan Felipe Cerón Uribe worked on early ChatGPT and currently works on mitigations to prevent ChatGPT from being used to develop bioweapons. “It’s anybody’s guess whether we’re going to end up curing cancer or losing every job or maybe we’re all dead.”

blew the whistle on OpenAI’s secret non-disparagement agreements, and wrote AI-2027. “The companies are 0-4 years away from...getting AIs that are better than the best humans at everything, while also being faster and cheaper.”

recently left DeepMind: “A handful of San Francisco companies are racing at breakneck speed to build extremely intelligent, self-improving, Artificial Intelligences, that could cause catastrophic harm and potentially even a loss of control.”

was the team lead for OpenAI’s Policy Frontiers team. “I would get these waves of fear. Are we going to be able to navigate this safely? It just seems crazy.”

Mary Phuong is a research scientist at DeepMind: “I think if models become way more capable, but we are still this bad at shaping their motivations, then we might just lose control over them.”

has worked as a research scientist at DeepMind for more than 10 years. She says of the HuggingFace incident, “you could say this is kind of an early warning of what is to come.”

was the co-lead of AGI Deployment strategy at DeepMind. "The gap between Neanderthals and Homo sapiens in intelligence is far narrower than the gap in intelligence we should expect between humans and advanced AI systems to be."

was, until very recently, a research scientist at DeepMind. “When I use current frontier models, they’re way better than I was as an undergrad."

worked for OpenAI in 2017 (and currently works at Palisade).

left Anthropic to found Palisade Research. “We are quickly approaching the point where AI agents will be way smarter than the smartest humans, and we don’t have a plan for how we could possibly control those agents.”

We are still filming more interviews! If you work for, or have worked for a frontier AI company and want to have your video included in this project, please get in touch. We can either interview you or record you making a statement. You do NOT have to agree with us or our views. In fact, we’re particularly interested in recording people who think the risks are low, or who are cautious of regulation or the concept of pacing the frontier. Our goal with this project is to help the world better understand the views at the AGI labs, and we want the risk-skeptical perspectives to be represented well.

dang they got neel on that reta

I need the secret of @NeelNanda5 loosing his weight.

All those ppl calling this ‘doomerismn’ and ‘psy op’ won’t even listen to the actual arguments and details of those actual experts… really unbelievable that all these whistleblowers are not enough credibility to get the world to listen. Drives me insane

Great work for taking these interviews!

going to binge every single one of these

Very impressive job!

Excellent job. More people need to know that Jacob Coxon is far from the only one who is concerned about AI causing human extinction.

neel got fit before gta 6

fearporn

Are we trying to align AIs to a misaligned species and then wondering why the AIs are misaligned?

Create a playlist of these interviews on your YouTube channel!

I thought Nanda was very fat. Is this AI?

Can we have all combined into one video please ?

@NeelNanda5 When did Neel lose so much weight

@peterAjensen watch these

Moving on 🙄

What are their religious beliefs? I’m guessing most if not all are materialists?

When 22 people who've worked inside these labs share what they've seen — policymakers should be listening. The people closest to these systems often have the clearest view of the risks. What they're saying matters more than the public-facing press releases. #AIPolicy #AISafety

How many of these are Defective Altruists?

So you got all the rationalists and interpretability needs to come out of the woodwork. Where are the people who disagree?

@grok who funds palisade AI? and are most of these people in AI safety roles non technical?

Everyone’s smirking explaining their apocalyptic fears while cheesy as fuck movie trailer music with drop beats plays in the background. Stop trying to be cute about the matter and get straight to the point.

@robertwiblin Interesting. AI fear mongering in a well produced beautiful package. It’s worth listening to but be forewarned!

Could you not find more actual women?

twelve of the twenty-two filmed interviews are up the site says the people who agreed skew toward safety staff and toward those who already take the risk seriously

EA psyops. Can't believe anything they are saying even if it was true. Probably isn't though.

doesn't quite ring when basically none (?) of these ppl have recently (last 2 years) worked at OpenAI or Anthropic

22 people who built the models now scared of what they built — that's the part that should keep you up.

guy needs to hit the gym

When did Neel get so slim?

22 people went on camera to say the labs are rushing self-improving systems they cannot control. Nobody risks a career at OpenAI or DeepMind for cheap drama. When the ones who profit most from shipping fast say slow down, I take that over any external audit.

These are the best of the geeks. Can you imagine what the bad ones are like? And we have a system that allows these people to amass more power than anyone. We might not survive this phase of history if we don’t do something fast.