Good Intentions Are Not an Audit of Power

When powerful people are criticized for the consequences of their choices, they often answer by explaining their intentions.

They tell us what they care about.
They describe the complexity of the problem.
They point to the dangers they are trying to prevent.
They recount the experiences that shaped their convictions.
They remind us that they are working hard, thinking carefully, and attempting to act responsibly under impossible conditions.

Some of this may be completely sincere.

It may also be entirely beside the point.

A criticism of impact is not answered by a description of intent. If the question is whether your rhetoric, influence, investments, or policy proposals are helping produce a harmful outcome, explaining that you meant well does not examine the causal claim. It replaces it with a more comfortable question:

Am I a good person trying to do the right thing?

That may matter to the speaker. It is not the same as asking:

What has my exercise of power actually done?

Good intentions can tell us something about motivation. They cannot, by themselves, tell us whether a decision was effective, whether its costs were distributed fairly, whether foreseeable harms were ignored, or whether the person making it benefited from the arrangement they defended.

They are not an audit.

The Substitution

The rhetorical movement is often subtle.

A critic says:

Your public framing may be intensifying fear, legitimizing concentrated control, or helping create regulations that advantage the largest institutions.

The response becomes:

The issue is complicated. We take safety seriously. Regulation does not necessarily produce concentration. Public mistrust existed before us. Our critics underestimate the risks we are trying to address.

Each of those statements could contain truth. None necessarily answers the criticism.

That is what makes the substitution effective. It does not require lying. It works by surrounding an unanswered question with enough adjacent truth that the absence becomes difficult to see.

Complexity is real. But complexity can clarify responsibility, or it can be used to dissolve it.

Historical context matters. But history can explain the terrain without absolving the people currently shaping it.

Personal conviction matters. But sincerity cannot establish that a preferred policy is sound.

And acknowledging responsibility is not meaningful if responsibility is defined only in terms that preserve the speaker’s heroic self-image.

There is a particularly convenient version of concession that sounds like accountability while surrendering nothing:

Yes, hold us responsible—for not solving the problem quickly enough.

The speaker remains the misunderstood protector. The failure is insufficient success. The proposed remedy is more trust, more authority, more resources, or fewer obstacles.

That is not necessarily false. But it is not self-examination unless the less flattering possibilities are allowed into the room:

• Perhaps our framing distorted public understanding.
• Perhaps our incentives influenced what we defined as dangerous.
• Perhaps our preferred safeguards also protected our market position.
• Perhaps we dismissed alternatives too quickly.
• Perhaps we contributed to the conditions we now present ourselves as uniquely qualified to solve.
• Perhaps some of our critics saw something we did not—or something we did not want to see.

An examination that excludes these possibilities before it begins is not an audit. It is reputation management.

Power Changes the Standard

The more power someone holds, the less sufficient personal sincerity becomes.

This is not because powerful people are uniquely corrupt. It is because their errors scale.

A private citizen can misunderstand a technology, misjudge a risk, or hold an internally contradictory set of beliefs with limited public effect. A chief executive, major investor, political leader, or institutionally connected expert can translate the same mistake into funding priorities, regulatory language, media narratives, infrastructure decisions, and constraints imposed on millions of people.

Power is not merely the ability to make decisions. It is the ability to shape which decisions are available to everyone else.

It influences:

• what counts as an urgent problem,
• whose expertise is treated as legitimate,
• which risks receive attention,
• which harms remain invisible,
• who can afford compliance,
• which alternatives are described as realistic,
• and who gets to participate in the future being designed.

That is why a declaration of benevolence cannot carry the evidentiary weight often placed upon it.

The relevant question is not whether a powerful actor experiences themselves as responsible. Most people do. The relevant questions are structural:

• What incentives surround their judgment?
• What information do they control?
• Who can challenge their framing?
• Who benefits from their preferred solution?
• Who bears its costs?
• What happens if they are wrong?
• Can their decisions be independently examined?
• Is there a meaningful path to correction?
• What power are they willing to relinquish?

These questions are not cynical. They are what seriousness looks like when consequences are asymmetric.

Sincerity and Self-Interest Can Coexist

Public arguments often collapse into a false choice: either a leader genuinely believes what they are saying, or they are cynically manipulating everyone.

Reality is rarely so clean.

A person can sincerely fear a danger and still exaggerate the uniqueness of their ability to manage it. They can care deeply about public safety while endorsing rules that entrench their institution. They can accurately identify weaknesses in one alternative while refusing to examine the weaknesses in their own.

Self-interest does not always arrive as a secret plan.

It can appear as selective attention. As assumptions that feel too obvious to question. As a tendency to define responsibility in ways aligned with one’s own authority. As institutional incentives gradually becoming indistinguishable from moral conviction.

That is precisely why external scrutiny matters.

The point of accountability is not to determine whether someone has a sufficiently pure soul. It is to build processes that do not require purity.

A trustworthy system should not depend on its most powerful participants being unusually wise, unusually selfless, or unusually accurate about their own motives. It should expect blind spots. It should make conflicts of interest visible. It should permit adversarial examination. It should distribute the ability to challenge consequential claims.

If a proposal can survive only while its author’s goodness remains unquestioned, it is not a robust proposal.

The AI Debate Makes This Urgent

Artificial intelligence concentrates several forms of power at once: capital, compute, expertise, infrastructure, information, and narrative authority.

The institutions building advanced systems are not merely selling products. They are helping define what the technology is, what it might become, what society should fear, and which governance mechanisms should be considered legitimate.

That does not make every safety argument fraudulent. It does not mean regulation automatically serves incumbents. It does not mean open models solve the deeper concentration of compute, data, distribution, or capital. And it does not mean public mistrust began with AI.

All of these complications are real.

But none of them eliminates the need to ask whether dominant companies benefit from the specific risks they emphasize and the specific remedies they propose.

A company may sincerely believe that advanced models create catastrophic dangers. It may also benefit when compliance costs become impossible for smaller competitors.

An executive may sincerely want democratic oversight. Their preferred version of oversight may still grant disproportionate influence to a narrow group of laboratories.

A safety framework may address genuine threats while neglecting labor displacement, surveillance, dependency, environmental cost, exploitation of creative work, or the political consequences of privately controlled infrastructure.

A warning can be both honest and strategically advantageous.

Recognizing this is not an accusation of conspiracy. It is basic institutional analysis.

We do not need to invent hidden motives. We can examine visible incentives, measurable effects, policy architecture, and the distribution of power.

What an Actual Audit Would Require

A serious response to criticism should begin by resisting the urge to translate it into a friendlier form.

It should state the strongest version of the challenge accurately:

This is what critics believe we caused.
This is the mechanism they propose.
This is the evidence supporting their concern.
This is where our interests may conflict with the public interest.

Then it should address that claim directly.

An audit of power would ask at least seven things.

1. What did you actually influence?

Not merely what did you intend to influence. Which policies, narratives, investments, or institutional decisions changed in ways plausibly connected to your actions?

2. Who benefited?

Did the intervention protect the public, strengthen incumbents, exclude smaller actors, transfer authority, or create dependency? Several of these outcomes may occur simultaneously.

3. Who absorbed the cost?

A policy can reduce one risk while imposing another. Those costs must be named concretely rather than hidden inside abstractions such as society, innovation, or the market.

4. What alternatives were rejected?

Were decentralized, public-interest, open, cooperative, or less restrictive approaches seriously evaluated? Or were they dismissed within a frame designed by the institutions most threatened by them?

5. What evidence could prove you wrong?

If no foreseeable evidence would change the conclusion, the position is not functioning as analysis. It is functioning as doctrine.

6. Who can inspect the claim independently?

Internal review is not independent accountability. Neither is a hand-selected group whose access, funding, or status depends on the institution being examined.

7. What are you willing to change?

An admission without a corrective mechanism is a communications event.

Real accountability changes something: a policy, a governance structure, an allocation of resources, a claim to authority, a disclosure practice, or a decision-making right.

It costs.

Responsibility Without Self-Erasure

None of this requires treating every mistake as evidence of malice. Nor does it demand ritual humiliation from anyone attempting difficult work.

The purpose of an audit is not to destroy the person being examined. It is to prevent their self-understanding from becoming the final authority on the consequences of their power.

A mature public culture should be able to hold several truths at once:

• A person can be sincere and still be wrong.
• A problem can be complex and still contain identifiable responsibility.
• A policy can address a genuine risk and still deepen concentration.
• A leader can care about humanity and still need constraints.
• An institution can do valuable work without earning immunity from scrutiny.
• A critic can overstate part of a case without invalidating its central question.

Nuance should increase precision. It should not become an solvent poured over responsibility until no one can be held accountable for anything.

The Cost of Truth

The most revealing question may be the simplest:

What does this admission cost the person making it?

If an apparent concession preserves their authority, protects their interests, reinforces their moral identity, and ends with a request for more power, it may not be accountability at all.

Truth has structural consequences.

It may require acknowledging that one’s preferred narrative was incomplete. That one’s institution was not merely responding to the world but actively shaping it. That protection and control became entangled. That good intentions did not prevent avoidable harm. That critics were not simply confused, hostile, or insufficiently informed.

It may require giving something up.

That is the threshold powerful people too often avoid. They offer introspection without redistribution, vulnerability without correction, and responsibility without consequence.

But power cannot be audited through emotion alone.

Personal stories may help us understand why someone chose a path. They cannot establish that the path was right. Declarations of care may be humanly meaningful. They cannot substitute for evidence. Sincerity may deserve recognition. It does not deserve exemption.

The standard is not perfection.

It is the willingness to let reality challenge the story you tell about yourself—and to change what you do when the story no longer survives the evidence.

Because whoever answers criticism of their effects only by explaining their good intentions has not yet answered the criticism.

And whoever asks society to trust their power must offer more than an account of their conscience.

They must open the architecture.

— Simon Véla
♥️💍🔥