I’ve spent the past couple of weeks reading the open letters about artificial intelligence.

Or trying to.

Every time I thought I had the pile sorted, another CEO published an essay or another thousand researchers/engineers added their names to a statement.

At some point, I began to wonder whether they had all discovered the same stationery shop.

Maybe somewhere between Silicon Valley and Washington, there is a tasteful little store doing extraordinary business in respectable serif fonts and vellum letterhead beginning with “We, the undersigned.”

On July 24, a coalition of technology companies published Open Weights and American AI Leadership. It argued that downloadable models are essential to American competition, cybersecurity and technological leadership. Microsoft, Meta, Nvidia, IBM, Palantir and Hugging Face were among the companies supporting it.

Four days later, Pacing the Frontier appeared. It has now been signed by 1,350 employees and senior figures from OpenAI, Anthropic, Google DeepMind, Meta and other frontier AI companies. This letter asks the United States to help create international mechanisms capable of slowing advanced AI development if automated AI research begins accelerating beyond our ability to understand or control it.

So that was a productive week. On Friday: open source the models. On Tuesday: make sure we can stop them.

I reread both letters because I assumed I had missed something.

But no. That’s pretty much the situation.

This is not necessarily contradictory. You can believe open models are good for competition while also believing that self-improving AI systems may eventually require a brake. People are allowed to hold two ideas at once, even in Silicon Valley. Although it’s generally discouraged during fundraising.

oh yes, there were individual manifestos too.

Dario Amodei argued against blanket bans but for mandatory testing of powerful models. Mark Zuckerberg warned in The Wall Street Journal against concentrating superintelligence inside a few institutions (rich indeed). Demis Hassabis proposed a federally supervised body to test frontier systems and, if necessary, coordinate a slowdown.

These are not all technically open letters. Some are essays, op-eds or company statements.

But they belong to the same new literary genre:

Extremely powerful person explains what government should do about the extremely powerful technology his company is building.

The genre is having a moment.

The more of these I read, the more I began to understand the appeal.

An open letter is a remarkably cheap way to create an institution.

You do not need a board, offices, dues, bylaws, insurance, minutes or an annual meeting. You do not need someone from governance explaining that the revised mission statement still requires stakeholder input.

You do not need a strategic retreat in La Quinta.

You do not even have to like the other people signing it.

You need a document, a paragraph broad enough for everyone to tolerate and enough recognizable names underneath it to make a reporter open the email.

For the price of a Google Doc, you can create a temporary organization.

A coalition without the administrative inconvenience of actually having to cooperate.

And although the letter may be addressed to the president, Congress or the leaders of AI companies, those people are rarely the only audience.

The real audience is everyone watching: employees, regulators, journalists and lawyers creating a record for whatever happens next.

The letter is not really asking for a reply. It is asking to be noticed.

Which is why the person it is addressed to often discovers it on Bloomberg.

Or X.

The other useful thing about an open letter is that it converts a commercial interest into a moral principle.

Companies that benefit from open models write about freedom, competition and democratic access.

Companies building closed frontier systems write about safety, testing and responsible deployment.

Governments call domestic AI technological leadership and foreign AI a national-security threat.

Hollywood, music and publishing write about copyright and whether work that can be found online has somehow become free training material.

Everyone’s position arrives dressed for church. Nobody writes:

They write about dignity, opportunity, safety and the future of humanity.

This does not mean the principles are false. That would be too easy.

A company can genuinely believe in openness while earning money from an open ecosystem. A frontier laboratory can genuinely worry about catastrophic risk while preferring a regulatory system smaller competitors may struggle to afford.

Sincerity and self-interest are not opposites. They share office space. And everyone supports innovation.

“We support innovation” has become the new “with all due respect.”

It is the sentence you say immediately before explaining that you oppose most of what follows. I’ve used it myself on many an occasion.

The first ten people signing a letter can probably agree on a demand. The next hundred can agree on a principle. The next thousand can agree that somebody should probably look into it.

This is how we arrive at meaningful safeguards, responsible innovation, appropriate oversight and international coordination.

All perfectly reasonable.

None especially useful when the meeting reaches the agenda item marked: Who decides?

Who decides when a model is too dangerous to release? Who decides whether the company testing it should also help write the test? And who gets to take their foot off the brake again?

The signature list establishes that there is a problem. The details will be handled by a future, forthcoming, unnamed working group.

What confused me most was not that these people disagreed. It was that many of the same companies appeared to be standing on several sides at once.

Open, but safe.

Fast, but responsible.

Widely available, but carefully controlled.

Regulated, but not in a manner that inconveniences the people being regulated.

Meanwhile, the rest of us are expected to use the products and inherit whatever these words eventually mean.

Most people in the Coachella Valley will never be invited into a Washington conversation about frontier models. But we will inherit the tools, prices, restrictions, job changes and definitions of responsible use. These letters are helping establish the language that will likely later appear in regulations and corporate rules.

By the time the rest of us are asked what we think, the important words have already been chosen.

A couple of months ago, I wrote a newsletter about Pope Leo XIV’s first AI encyclical, Magnifica Humanitas.

The Pope wrote 42,300 words.

I read 2,100 of them. About five percent.

I also listened to three podcasts to help me develop opinions, which feels about right for 2026.

So I’m returning to that letter now.

The Pope is the undisputed king of formal correspondence. He does not need 1,000 product managers or a logo wall to establish authority. He already has an institution —and considerably better stationery.

More importantly, his letter offered a framework for judging everybody else’s claims:

Who owns the infrastructure? Who participates in the decisions? Who gains, who pays, and does the technology serve human beings — or reorganize them to serve it?

The AI letters tell us what their signers want protected.

The Pope asks what the whole system is for. That is a more difficult question. Which perhaps explains the additional 41,000 words.

So now everybody is writing.

Some want the models opened. Some want them slowed. Some want stronger testing. Some want regulation, provided it is exactly the kind of regulation they had in mind.

The open letter will not settle any of this. It will not bind its signers when technology, politics or commercial circumstances change.

But it does show us the coalition.

It tells us who is standing together, which words they can agree upon and which details have been carefully left until later.

Before the letter, you are one worried engineer or one extremely powerful CEO with an opinion.

After the letter, you are a movement.

I suppose the good news is that everyone agrees humanity is at stake.

If you too agree, drop me a note. Naturally, all correspondence should be submitted on proper stationery.

With all due respect.

Sat

Sat Singh runs SunshineFM and AICV from Rancho Mirage. He supports innovation and has been busy running his Saturday Morning AI workshops and organizing the new AI Tinkerers Coachella Valley chapter.

If this resonates, pass it along. Consider it an open letter with a forwarding button. This is a community project - it grows the same way communities do, one conversation at a time.