US asks China to share alerts when AI goes wrong

America and China both build very smart computer programs called AI. On Sunday, America said: if one of ours does something dangerous, we’ll tell you, and you tell us. It’s like two neighbors agreeing to shout if either one’s grill catches fire. It’s only a plan so far, and the grown-ups have it.

Why it worksA five-year-old reasons from what is in front of them, so the script trades the notification mechanism for two neighbors who agree to shout about a fire, which is what Bessent described. Expect “is our computer going to be bad?” Nobody has named one that is. This is the grown-ups agreeing to tell each other.

Also written for 8–12 and 13–17.

The US and China are racing to build the smartest AI, and neither can see what the other is building. On Sunday the US proposed a fix: if an AI incident on one side could threaten national security, that side tells the other. The buddy check at the pool: you swim your own race, but you look over.

Why it worksThis age is working out cause and effect and whether a rule gets applied both ways, and this deal is a rule that only works if both sides keep it. The buddy check is the mechanism: you race, and you still look over. Expect “what if China lies?” Then it doesn’t work. Nobody has signed anything; it was discussed on Sunday.

Also written for 5–7 and 13–17.

If the US and China are racing to win at AI, why would either tell the other when its AI misfires? Bessent’s answer on Sunday: the top two AI powers should be less opaque, so the US proposed an alert system for incidents that touch national security. Can you race someone and share crash reports?

Why it worksA teenager can hold two true things at once, and the script hands them the pair: a race to win at AI, which Trump says is the only thing that counts, and a promise to tell the rival when American AI goes wrong. The last question is real. Expect “would China actually tell us?” Say you don’t know. Nobody does yet.

Also written for 5–7 and 8–12.