more computing power goes into training each new top AI model, every year, since 2020.
More computing generally means more capable systems.
Research — Advanced intelligence & human life
The promise and the peril of advanced intelligence, recorded by the people living it.
Civilization Signal is a systematic response in preparedness ahead of the AI/superintelligence transition. It is longitudinal public participation research on AI/SI safety and its impact on humanity: the breakthroughs that heal, teach and connect us, and the harms that must be prevented.
Ask the people building AI, vote on what matters most, answer a daily question, share a claim or a solution. Everything joins an ongoing public record that informs policymakers and helps nurture superintelligence with human wisdom and nobility.
Write or say a question for AI leaders. Pick who should answer, or leave it open. Opening soon
Vote for the questions you want answered. The most-voted questions make it to the next edition: SUMMIT 333 AI/SI Safety, Davos, 19 January 2027. A vote shows what people want asked. It does not decide what is true. Opening soon
A daily question that rotates every Monday, picked from the week's news and what people voted for. Or tell us what AI has changed for you. Opening soon
Tell us what AI has changed in your life or work, or an idea that could help. Opening soon
What is shared goes into an open archive that researchers, policymakers and AI builders can read.
The clock started in April 2026, with 18 months to get AI/SI right. Eleven are left.
Countdown code · to get AI/SI safety right
18
Why take part now1 / 6
The pace, in numbers
more computing power goes into training each new top AI model, every year, since 2020.
More computing generally means more capable systems.
is how often the length of tasks AI can finish on its own doubles. It has held for six years.
At that pace, the jobs AI can do alone keep getting longer.
of the world’s people used generative AI within three years of its launch.
This is why real lives, not just data, need to be part of the record.
more documented AI incidents in 2025 than in 2024: 362, up from 233.
These are public reports of AI causing harm or failing. More use means more that can go wrong.
Why take part now2 / 6
In their own words
“Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.”
people have signed a call to stop building superintelligence until scientists agree it can be done safely and the public backs it.
Why take part now3 / 6
Why it takes all of us
That makes this a volume game. We need everyone on board, so sharing is not a nice-to-have. It is the work. The more people share, the sooner leaders and AI stay in touch with real life.
Why take part now4 / 6
The code of SUMMIT 333
299.792° 18′ 175″
As of April 2026
It started at 18 in April 2026 and runs out in October 2027.
Move at the speed of light. This is our motto, and these three numbers are our code.
Light travels 299,792 kilometers every second. It is how fast the world is changing, and how fast we must learn to act.
The window opened in April 2026 with 18 months and closes in October 2027. The big number above falls by one every month.
What we decide now shapes seven generations, about 175 years. The future remembers.
Why take part now5 / 6
Four reasons to start today
The EU AI Act entered into force on 1 August 2024, and in 2026 the EU pushed its high-risk rules back to December 2027. Other countries are drafting theirs. Choices made in this window are hard to undo. Source: European Commission ↗
About 2.2 billion people are still offline, so their lives are missing from the data AI learns from. Every person who joins helps close that gap. Source: ITU, 2025 ↗
Change over time can only be shown if we begin counting now. Each month without your voice is a month the record cannot get back.
When voting opens, the most-voted questions go to SUMMIT 333 AI/SI Safety in Davos on 18–19 January 2027. Vote before then to have yours counted.
Why take part now6 / 6
Every question is reviewed before it appears. A question that names a real person is held for extra review.
A vote counts how many people want a question asked. A vote never decides what is true, and it never changes a question’s status.
The most-voted questions are put to leaders at SUMMIT 333 AI/SI Safety in Davos, 18–19 January 2027, and at the editions after it.
Every question carries a visible status: submitted, shortlisted, assigned to a session, answered. If a leader does not answer, the record says so.