“We should not train models that we cannot make an extremely strong case they will be able to keep under human control.”
AI could be the best thing we ever built - or the worst.
This year OpenAI’s AI agents twice broke out of the systems built to contain them. In September, OpenAI and Anthropic told the UN they will slow down.
“No leader, no company, and no nation can manage this alone.” - Dario Amodei, Anthropic, at the UN
Voices · from inside the industry
They agree on two things. It could be huge. It needs rules.
The heads of the companies building AI and the scientists who pioneered it. Every quote links to its source.
“Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.”
“One day maybe we can cure all disease with the help of AI.”
“The world needs to establish the rules of the road so that any downsides of artificial intelligence are far outweighed by its benefits.”
“We’ve never had to deal with things more intelligent than ourselves before.”
“We think that regulatory intervention by governments will be critical to mitigate the risks of increasingly powerful models.”
“We, as the producers of this technology, have a duty and an obligation to be honest about what is coming.”
“Should we risk loss of control of our civilization?”
“A single pathogenic experiment could spark a pandemic, a tiny molecular event with global ramifications.”
#01 / FIRST, THE STAKESTODAY
First, the stakes. The same AI could heal, feed and teach.
Earlier diagnosisAI catches what’s easy to miss
New medicinesAn antibiotic discovered with machine learning
Clean energyCheaper solar, wind and batteries
FoodCrops that survive the heat
Early warningsFloods and fires predicted earlier
SchoolSupport for teachers
#02 / WORKTODAY
AI does it faster and cheaper. Who can resist?
01A person works remotely
02AI does it faster
03Companies feel the pressure
#03 / MONEY AND THE STATEPOSSIBLE SCENARIO
The state runs on your work. What if it’s gone?
01Work pays for what we share
02The budget shrinks
03Services crack
#04 / SPEED AND CONTROLTODAY
AI is speeding up. It has already slipped out of a test.
01An open cage
02The swarm
03Searching for clues
#05 / WHO DECIDESEXTREME SCENARIO
What if we’re not needed?
01AI trains robots
02Robots in every job
03People on the sidelines
#06 / THE RULESA WAY FORWARD
Rules before the harm scales up.
01Companies race ahead
02The state plays catch-up
03Who is responsible?
07 / WHAT CAN I DOA WAY FORWARD
Don’t stop progress. Set its limits.
You don’t buckle up because you know there will be a crash. You buckle up because you don’t know.
86% of Americans want AI companies held to independent safety standards, even if it slows AI down. Quinnipiac, 2026 ↗
What good rules look like
- Independent testing before AI reaches millions of people
- Clear responsibility when an AI system causes harm
- AI’s gains shared fairly - through taxes and a real safety net
What you can do
- Share this with one person who hasn’t thought about it yet
- Ask the people who represent you where they stand on AI rules
- When AI companies publish safety plans, read them - and ask questions