← ELI5 中文
GET IT AT A GLANCE

The White
House AI
lunch

The table of people who build the most AI,
who signed a piece of paper after eating
No. 38How AI WorksPart 7 of 14
1

The short version

On September 29, 2026, the US president invited a table of America's top AI bosses to lunch at the White House. After eating, they signed a piece of paper: "We will keep ourselves in check."

The same day, the president also renamed AI: in government documents it will no longer be called "artificial intelligence." It will be called "superintelligence."

2

How is this
different from
usual?

USUALLY
When something new is dangerous, the government usually sets the rules
Like needing a driver's license: the examiner isn't you.
THIS TIME
The government said: no rules from us, you police yourselves
The reason in one line: "We're racing China, and we can't slow down."

Musk's own description is the most vivid: "It's basically grading each other's homework."

3

What the paper
says

Using homework as the comparison, there are roughly four checkpoints:

1. Watch it yourselfKeep checking while it's being built: don't let it pick up bad habits or touch what it shouldn't
2. Classmates cross-checkInside the company, a separate team checks whether "the checking actually works"
3. Ask outsiders to grade itOutsiders do the checking, though the company picks which ones
4. Get a parent's signatureThe results go to the company's own board, which makes sure fixes happen

Seven people signed: the president, Musk, Zuckerberg, Jensen Huang, Anthropic's Amodei, Google's Pichai, and OpenAI's number two, Brockman.

4

What each person
at the table said

T
Trump · President"Whoever wins superintelligence wins. There won't be a second place."
M
Musk"The most likely outcome is high income for everyone." Pressed by a reporter on what happens with no jobs, he said: "Jobs have always been changing."
H
Jensen Huang · chipmaker"Safety and progress don't conflict." He also said the paper "has teeth."
Z
Zuckerberg"This is only the beginning."
A
Amodei · AnthropicThe only one at the table who spoke plainly:
"The risks are very real, and how to manage them hasn't been settled."

One paper, four signers, four different messages.

5

What they're
really anxious
about

US China ≈ 3x

Not safety. Electricity.

After lunch Musk and Huang talked for 26 minutes, and most of it was about where to find more electricity.

Musk said China generates about 3 times as much electricity as the US. Their answer is to build more power plants, and even to send solar panels and computers into space.

6

What this paper
is not

It is not a law Fail to follow it and nobody gets fined. It says "should," not "must." The president's own words: "morally binding."
The results of the checks are hidden from outsiders. Reports go to the company's own board, not to the government, and are not made public.
There are critics too. A senator's words: "The people building AI are themselves saying the technology is outrunning the safeguards, and the president's response is to change the name and let them police themselves."
"What does this have to do with me?"
For the near term, nobody is hitting the brakes on AI. The AI tools on your phone will get stronger and cheaper.

Whether or not everyone ends up with a "high income," the ones who lose out first in these years are the people who refuse to learn to use it.
UP NEXT

Tech Launch-Event Jargon, Decoded

Why "announced on stage" and "you can use it today" are months apart: the yearly tech events, the launch ladder, and how to read the fine print.

No. 39 · 6 minKeep going →
← Back to ELI5