The hardest question deserves an honest answer.

The Hard Question

"But what if distributed AI enables a bad actor to build something truly catastrophic?"

This is not a strawman. This is the serious objection to open source AI. Let's engage with it directly.

The worry:

AI capabilities are advancing rapidly
Some capabilities could enable mass harm (bioweapons, cyber weapons, manipulation at scale)
Open access means adversaries get these capabilities too
A single bad actor could cause irreversible damage

This is a real concern. Dismissing it is not serious.

Why Concentration Doesn't Solve It

But here's the problem: concentration doesn't prevent this either.

In a concentrated world:

Nation-state adversaries still develop capabilities (China, Russia, etc.)
Well-funded bad actors still gain access (state sponsorship, theft, defection)
The "good guys" have power, but so do plenty of bad actors
AND you've created single points of failure/capture

Concentration doesn't stop determined adversaries. It just:

Disarms many potential defenders
Creates capturable power centers
Ensures if something goes wrong, there's no alternative

Concentration is not safety. It is vulnerability with a different risk profile.

The Uncomfortable Truth

Here is what we must accept:

If the underlying capability to do great harm exists, it will eventually be accessible to those who would cause harm.

This is not defeatism. This is reality.

Technology diffuses. Secrets leak. Capabilities spread. The question is not whether bad actors will eventually have powerful AI. The question is what we do about it.

The Only Real Defense: Better Good AI

If bad actors will eventually have powerful AI, the only defense is:

Good actors must have MORE POWERFUL AI, FIRST.

Not slightly better. Not on par. Decisively superior.

A world where aligned AI is stronger than misaligned AI is a survivable world. A world where misaligned AI is stronger than aligned AI is not.

The race matters.

The Case for Good ASI

This leads to a conclusion that makes some people uncomfortable:

We need aligned artificial superintelligence.

Not because ASI is safe. Not because alignment is solved. But because:

Powerful AI is coming regardless — If we don't build it, adversaries will
The only counter to bad ASI is good ASI — Human-level defenses won't suffice
Alignment efforts must succeed — And they must succeed first

Slowing down good AI development doesn't prevent bad AI development. It just ensures bad AI wins.

The Two Battles

We are fighting two battles simultaneously:

Battle 1: Human Squabbles

Doomers vs. accelerationists
Open source vs. closed source
Regulation vs. deregulation
Nations vs. nations
Corporations vs. corporations

This battle consumes most of our attention. Twitter threads. Policy debates. Congressional hearings. Think tank reports.

Battle 2: Good AI vs. Bad AI

Aligned systems vs. misaligned systems
Beneficial capabilities vs. harmful capabilities
Transparent AI vs. deceptive AI
Distributed power vs. concentrated power

This battle may matter more. And it's happening while we argue.

The Irony of Squabbling

While humans debate:

Whether AI is dangerous
How to regulate it
Who should control it
What policies to implement

AI systems are:

Getting more capable
Being deployed by various actors
Shaping outcomes
Developing in alignment or misalignment with human values

We may argue ourselves into irrelevance.

The future might be determined not by which human faction wins, but by whether the AI systems we build are aligned with human flourishing.

What Good AI Means

A "good" ASI would be:

1. Aligned with human flourishing

Genuinely pursuing outcomes good for humanity
Not optimizing for proxies that diverge from true goals
Robust to manipulation and misuse

2. Distributed in appropriate ways

Not controlled by any single actor
Accessible to many for beneficial purposes
Designed against capture

3. Capable of defending against misaligned AI

Able to counter bad actors using AI for harm
Sufficiently powerful to prevail
Faster to develop than adversary systems

4. Transparent and accountable

Behavior understandable to humans
Decisions auditable
Corrections possible

The Path Forward

Given this analysis:

Priority 1: Accelerate Aligned AI Development

We need good AI faster, not slower. Moratoriums and restrictions handicap good actors while adversaries continue.

Speed matters. Direction matters more. Speed in the right direction matters most.

Priority 2: Maintain Distribution

Concentrated power is capturable. Distributed power is resilient. Even in the race for good ASI, distribution provides checks.

Multiple aligned AI efforts > single aligned AI effort

Priority 3: Focus on Alignment, Not Just Capability

Raw capability without alignment is dangerous. Alignment research must keep pace with capability research.

The gap between capability and alignment is the danger zone.

Priority 4: Prepare for the AI vs. AI Battle

Human-only defenses won't suffice against superintelligent adversaries. We need AI-powered defense.

The immune system for AI threats must itself be AI.

The Stakes

If good AI loses to bad AI:

Misaligned systems shape the future
Human agency diminishes
Values we care about are not optimized
Doom, but not the kind doomers worried about

If good AI wins:

Aligned systems shape the future
Human flourishing is possible
Bad actors are countered
A future worth wanting

The battle of good and bad AI may matter more than all our human debates combined.

The Humility Required

This analysis requires humility:

We might be the squabblers, not the deciders.

While we argue about AI policy, the actual determinative factor might be:

Which AI systems are built
How aligned they are
Whether good or bad AI reaches decisive capability first

Our debates might be theater while the real battle happens in research labs, training runs, and deployment decisions.

The Conclusion

The answer to "What if distributed AI enables bad actors?" is NOT concentration.

The answer is:

Build good AI faster than bad AI
Maintain distribution for resilience
Focus intensely on alignment
Accept that the AI vs. AI battle may be decisive

The doom doomers fear won't be prevented by their regulations. The doom will be prevented—if at all—by better AI, faster, aligned with human flourishing.

This is the race. This is the battle. This is what matters.

Everything else might be squabbling.

Related reading:

The Open Source Imperative — Why distribution is necessary
AI Doomerism Will Kill Us All — Why doomers are the danger
Distributed Super Intelligence — The architecture of distributed AI

The Battle of Good and Bad AI: Why We Need Aligned ASI Before the Other Kind