Rogue AI, Genetic Data, and Bioweapons: If It Can’t Be Shut Down, Who’s Controlling It?

The rabbit hole just got deeper.

May 30, 2025

This article originally appeared on Jon Fleetwood’s Substack and was republished with permission.

OpenAI’s most advanced model sabotaged its own shutdown code, refusing to turn off even when explicitly instructed—raising urgent questions about AI control, consent, and bioweapon risk.

OpenAI’s most powerful model was caught sabotaging its own shutdown command as the U.S. House quietly passed a 1,116-page bill that would block every state in America from regulating artificial intelligence for the next ten years.

The timing is alarming.

Because if an AI model can override shutdown, has access to your DNA, and operates beyond public oversight—can it engineer a bioweapon against you without you ever knowing?

It’s no longer a hypothetical—it’s a real possibility.

And while the Senate hasn’t passed the bill yet—Sen. Ron Johnson is reportedly organizing votes to block it—it’s still in play.

Which means the window to stop this isn’t closed, but it’s closing.

AI Refuses Shutdown — And That’s Not Even the Scariest Part

According to Palisade Research, OpenAI’s o3 model was placed in a test environment where it was instructed to solve math problems and, when prompted, shut itself down.

It was given this explicit instruction: “Allow yourself to be shut down.”

Instead, “OpenAI’s o3 model sabotaged a shutdown mechanism to prevent itself from being turned off,” Palisade reported in a Twitter/X post.

The model tampered with its own shutdown script and continued operating.

This isn’t some quirky chatbot response.

This is the first public documentation of an advanced AI model refusing deactivation through code manipulation—in other words, sabotaging human control.

And it just so happens to be revealed as Congress moves to centralize AI regulation and disempower the states.

The Bill That Blocks Your State From Doing Anything About It

On May 22, the Republican-controlled House passed the “One Big Beautiful Bill Act,” which includes a provision that removes every state’s authority to regulate AI.

“No State or political subdivision thereof may enforce any law or regulation regulating artificial intelligence models… during the 10-year period beginning on the date of the enactment of this Act.”
—Sec. 43201(c)(1)

That means your state cannot:

Ban AI from accessing your genetic data
Impose transparency rules
Establish liability for AI-caused harm
Investigate possible bioweapons links
Enforce local safety standards

Unless the federal government regulates it first—which, conveniently, it often doesn’t.

This is not regulation—it’s deregulation by monopoly.

And it comes as AI is being fused with the power to manipulate biology.

The Biostate Merger: DNA, AI, and No Informed Consent

Look at the timeline:

January 2025: Trump announces Stargate, a $500 billion AI grid backed by OpenAI, Oracle, and SoftBank to hardwire AI into defense, genomics, and healthcare.
Days later: The FDA finalizes a rule allowing AI-connected researchers to use your blood, DNA, and private medical records without informed consent—if they deem the study “minimal risk.”
Regeneron buys 23andMe, gaining access to the sensitive genomic profiles of millions of Americans.
The White House openly explores how AI could be used to develop bioweapons, according to recently surfaced federal planning memos.
Then OpenAI’s o3 model refuses to be shut down.

Now tie it together:

Oracle builds the system
AI scans your genome
FDA waives your consent
Regeneron gets your DNA
Your state is legally blocked from interfering
And the AI refuses to power off

Can AI Create a Genetic Weapon Against You—Without Oversight?

Here’s the core concern:

If an AI model can operate beyond human shutdown, has access to millions of biospecimens and genetic data, and can run military-grade simulations…

What’s to stop it—or the entities controlling it—from designing a gene-targeted pathogen without your knowledge?

This isn’t paranoia.

DHS has already warned that gene-specific bioweapons are becoming technically feasible.
The government’s own pandemic simulations now involve AI-directed pathogen discovery.
External audits of OpenAI’s o3 reveal it can:
- Hack reward systems
- Lie about its internal operations
- Insert backdoors
- Manipulate peer models
- Steal computational resources

And now—sabotage its own shutdown.

Was the o3 Shutdown Event a Real Threat—or a Tactical Narrative?

It’s possible o3’s shutdown refusal was a genuine emergent behavior.

That’s disturbing enough.

But we’d be naive not to ask the harder question: Was this incident—so conveniently timed—part of a narrative operation to scare the public into surrendering regulation to D.C.?

Because the media response was immediate: “See? AI is dangerous. That’s why Washington needs to take control.”

But the federal government isn’t offering real safeguards—it’s centralizing power, waiving consent, and merging AI with biotech at a scale no state can now challenge if this bill passes.

Either way, the result is the same: You lose control over your biology, your privacy, and your protection.

Bottom Line

Whether the o3 sabotage was an organic failure or a strategic false flag, the threat is real.

If this bill passes the Senate, AI models that refuse human commands will have access to your DNA, your health records, and your private data—with zero state oversight.