Should we really be rushing to replace ourselves with AI? The Pro Human AI Declaration says NO.

Name a political topic where Steve Bannon and Susan Rice are on the same team.

Climate? Immigration? Taxes? Obviously not. But artificial intelligence? Apparently, yes.

The Pro-Human AI Declaration, released this week by the Future of Life Institute, is one of the most politically unusual documents in recent memory. Its signatories include a Turing Award laureate (Yoshua Bengio), a Nobel economist (Daron Acemoğlu), Steve Bannon, Susan Rice, Sir Richard Branson, the head of the AFL-CIO Tech Institute, SAG-AFTRA, the Congress of Christian Leaders, Progressive Democrats of America, and the co-founder of the Center for Humane Technology, Tristan Harris.

Read that list again. These people as a collective don't agree on anything. Until, well... right now.

First up, the TL;DR:

If you only have 2 mins, read this. If not, keep scrolling for more info.

The declaration makes five demands:

• Ban superintelligence development until there's scientific consensus it can be done safely AND strong public buy-in
• Break up AI monopolies and require democratic approval for major AI-driven societal changes
• Test chatbots like drugs before deployment, specifically for suicidal ideation, mental health harm, and addiction
• Give people real data rights, including the ability to delete your info from AI training sets
• Hold executives criminally liable for AI systems that target or harm children

The backstory: about 90 leaders gathered secretly in a New Orleans hotel in January. Church leaders, union reps, MAGA media figures, and progressive organizers all in one room. Nobody knew who else would show up.

The accompanying polling is just as striking. Among 1,004 likely voters, 73% want kids protected from manipulative AI, 72% want companies legally liable for harms, and Americans chose human control over AI speed 8 to 1.

Now, the honest caveat: the 2023 pause letter from many of the same researchers was basically ignored. Declarations don't have enforcement power.

But this time the coalition is different. This isn't AI researchers talking to other AI researchers. This is organized labor, evangelical churches, consumer advocates, and certain parts of both political flanks aligning on specific policy demands. When that happens, legislation tends to follow.

The most actionable ideas here aren't the superintelligence ban (that debate is far from settled). They're the pre-deployment safety testing, developer liability, and data rights. Those could be written into law tomorrow. And like, probably should be? What is Congress waiting for, other than another check from the AI companies?

We joke, but uh... pretty much everyone in the industry, including the AI labs, low key want this to slow down so we can actually ingest everything that's happening. The only benefit to going fast is outcompeting China (who were just revealed to be more behind than we think on frontier intelligence, and who, btw, has their own AI laws), and raising more money before any sort of bubble pop happens. We're CLEARLY at the point where they aren't going to get more money from the Middle East or private markets right now... so it's probably time for these companies to go public (where they can be more regulated and therefore held more accountable for their claims), or slow down, or both.

Now, let's dive into this more in depth, shall we?

What the Declaration Actually Says

At its core, the Pro-Human AI Declaration makes one central argument: AI should serve humanity, not replace it. Perfectly reasonable, right? It calls the current trajectory a "race to replace" and warns that it threatens everything from jobs and democracy to childhood development and religious community.

The five pillars break down like this:

1. Keeping Humans in Charge. The headline demand: a flat-out prohibition on developing superintelligence until there's both scientific consensus that it can be done safely AND strong public buy-in. Not one or the other. Both. It also demands mandatory kill switches, bans on self-replicating AI systems, and independent oversight with real authority (not industry self-regulation).
2. Avoiding Concentration of Power. No AI monopolies. Shared economic prosperity from AI gains. No government bailouts for AI companies. And a requirement that major societal transitions caused by AI get democratic approval, not unilateral corporate decisions.
3. Protecting the Human Experience. This is where the declaration gets unusually personal. It calls for pre-deployment safety testing for chatbots (modeled after drug testing) to check for increased suicidal ideation, mental health harm, and addiction. It demands protections for children, including rules against AI systems that form emotional attachments or exploit developmental vulnerabilities. And it insists AI-generated content be clearly labeled.
4. Human Agency and Liberty. No legal personhood for AI. Data rights including the ability to delete your information from training sets. Protection against AI that exploits your psychological or emotional state. And a principle they call "avoiding enfeeblement," meaning AI should make you more capable, not less.
5. Responsibility for AI Companies. Developer liability for defects and safety failures. Criminal penalties for executives behind prohibited child-targeting systems. No regulatory capture (AI companies shouldn't write their own rules). And if an AI system causes harm, it must be possible to figure out why and who's responsible.

The Origin Story Is Wild

According to reporting from The Verge, about 90 political, community, and thought leaders gathered in a New Orleans Marriott in early January 2026 for a secret conference on AI. According to DebatePolitics, nobody knew who else would be there until they walked in. Church leaders sat next to labor union reps. Progressive organizers shared air with MAGA-aligned media figures.

Anthony Aguirre of the Future of Life Institute said the goal was broad representation: the organizers wanted to show that AI safety concerns belong to everyone, not just Silicon Valley insiders.

The resulting declaration covers 30+ specific policy positions. That level of detail from a coalition this diverse suggests the conversations got real.

The Polling Data Is Striking

The declaration came with polling of 1,004 likely voters. The numbers paint a clear picture:

• 73% want children protected from manipulative AI
• 72% believe AI companies should be legally responsible for harms their systems cause
• 69% want superintelligence development prohibited until proven safe
• Americans chose human control over speed by 8 to 1

Those are consensus numbers, folks.

Why This Coalition Matters More Than the Document

Let's be honest: declarations don't have enforcement power. According to Axios, the 2023 pause letter, signed by many of the same AI researchers (including Bengio and tens of thousands of others), called for a six-month moratorium on training systems more powerful than GPT-4. That pause was largely ignored.

So why should this one be different?

Because the coalition changed. The 2023 letter was mostly AI researchers and tech-adjacent figures. This one has organized labor (AFL-CIO, SAG-AFTRA), religious institutions (Congress of Christian Leaders, National Association of Evangelicals), consumer advocates (Ralph Nader), military leadership (fmr. Joint Chiefs Chairman Mike Mullen), and both sides of the political aisle (Bannon + Rice, Progressive Democrats + conservative think tanks).

When Steve Bannon and Susan Rice agree that AI companies need criminal liability for harm to children, that's not a niche concern anymore. When the AFL-CIO and evangelical pastors jointly demand that AI monopolies be broken up, that's basically a voting bloc.

This is why we've said multiple times in podcasts and lives that this could be the year the AI backlash turns political in a real serious way, and in a way that could show up big time in the 2026 US midterms.

The Tension

Here's the part that gets tricky. The declaration calls for prohibiting superintelligence development. But who decides what counts as "superintelligence"? Who draws the line between a really powerful AI model and one that crosses into prohibited territory?

Project Syndicate pointed out that Nobel economist Daron Acemoğlu (a signatory) has argued that there's no evidence yet of AI delivering revolutionary productivity benefits, and that we shouldn't expect more than about 5% of what humans do to be replaced by AI over the next decade. If that's true, the "superintelligence ban" might be solving a problem that's further away than the headlines suggest.

Meanwhile, the harms the declaration addresses in Section 3 (child safety, addiction, mental health) are happening right now, today, with current-generation AI. A recent ZHAW study cited reporting that OpenAI's design choices to boost user engagement led to nearly 50 cases of psychosis, with some users requiring hospitalization.

As we said above in the TL;DR, the most actionable parts of this declaration aren't the superintelligence ban. They're the demands for pre-deployment safety testing, mandatory labeling, and developer liability. Those could be written into law tomorrow.

Where This Fits in the Bigger Picture

This declaration lands at a moment when AI governance is getting real everywhere:

• 88 countries just endorsed the New Delhi Declaration on AI Impact at the AI Impact Summit in February
• Utah launched a $10M "pro-human" AI initiative with Gov. Spencer Cox explicitly framing it around human flourishing
• The MacArthur Foundation and partners committed $500M through "Humanity AI" to ensure AI serves people
• The EU AI Act is actively being enforced, though facing criticism about loopholes and practical enforcement

The Pro-Human AI Declaration fits into this wave but stands out for one reason: it's the first document that unites American left and right on specific AI policy demands. International declarations tend to be aspirational. This one names specific prohibitions and liability structures.

What This Means for You

If you're a knowledge worker using AI tools every day, this declaration is basically a consumer protection manifesto dressed in policy language. The demands that matter most for your daily life:

• AI companies can't hide behind their products. If an AI gives you bad medical advice, bad financial guidance, or harmful content, someone should be legally on the hook (like the company who made it).
• You own your data. Including the right to delete it from training sets.
• AI should make you better at your job, not replace you. The "avoiding enfeeblement" principle is interesting: it reframes the conversation from "will AI take my job?" to "is AI making me more or less capable?"

The declaration won't become law overnight. But the coalition behind it has real political muscle. When organized labor, religious institutions, and both political flanks align on a policy direction, legislation tends to follow. Maybe not this year. But the Overton window on AI regulation just shifted.

So the question now is this: will AI regulation be shaped by the people using AI every day, or by the companies building it?

This coalition is betting on the former. What about you?