Our principles
We are asking you to trust a record. These are the rules we keep it by.
01
Show the record, not our opinion of it
Every claim on this site links to where it came from, so you can check us.
02
Plain language, always
If a sentence needs a degree to read, it is our failure, not yours.
03
Nobody is a villain in one line
People disagree in good faith. We show what each of them actually said.
04
Your privacy is not a price
We do not track you. Your own estimate of the risk is never saved.
The four above are rules we already keep. What follows is our wager — what we think should be asked of the people building this. You are free to disagree with it and still use the record.
05
Nobody should be able to build something this powerful in privateOur wager
A handful of companies are training systems they themselves say could be dangerous, and they decide alone what is safe to release. We think the tests, the results and the failures should be visible to people outside the company that ran them.
06
The people who carry the risk should have a sayOur wager
The costs of getting this wrong fall on everyone: on workers, on children, on people who never agreed to any of it. Decisions of that size belong to the public, not to a boardroom.
07
Speed is a choice, not a law of natureOur wager
The race is often described as inevitable. It is not. It is the sum of choices made by a small number of people who could choose differently, and who have said in their own words that they feel unable to slow down.
08
Anyone inside should be able to warn usOur wager
Employees at these companies have signed away their right to speak. We think anyone who sees a serious risk should be able to raise it without losing their livelihood, and that this protection should be written into law rather than promised in a blog post.
09
A promise with no consequence is not a safeguardOur wager
Most of the safety commitments on this site are voluntary. We track them anyway, because the gap between what was promised and what was done is the clearest evidence we have about whether self-regulation works.