On Intelligence, Evidence, and the Predicament of Accountability

By Andrew Klein

There is a difference between knowing something and being able to prove it. Most people understand this intuitively. The law understands it systematically. And the systems that produce the greatest harms are designed to exploit the gap.

I want to write about that gap — the distance between what we can know and what we can prove in a court of law — and why it matters more now than it ever has.

I. Two Kinds of Information

Intelligence is information that helps you understand what is happening. It comes from many sources: anonymous testimony, intercepted communications, pattern analysis, defectors, journalism, documentaries. Its purpose is to inform, to guide further investigation, to build a picture. It does not need to be admissible in court. It needs to be useful.

Evidence is information that can be presented in a court of law to establish facts. It has strict requirements. Chain of custody. Verifiability. The right of the accused to confront witnesses. The ability to cross-examine. Anonymous testimony, by definition, cannot be cross-examined. The defence cannot question the witness. The identity of the witness cannot be verified. The context of the testimony cannot be tested.

This distinction is not a technicality. It is the difference between understanding and accountability.

Intelligence tells us what is probably happening. Evidence establishes what can be proven beyond reasonable doubt. Both are necessary. Neither is sufficient alone.

II. The Predicament of Anonymous Testimony

Consider a documentary that presents testimony from anonymous military and intelligence personnel. The sources describe targeting systems that reduce human beings to data points. They describe decisions made by algorithms with limited human oversight. They describe the anticipated number of civilian casualties being calculated before strikes are approved.

The anonymity makes sense. The sources fear reprisal. If their identities were known, they would face prosecution, imprisonment, or worse. The anonymity is a condition of the testimony existing at all.

But that same anonymity makes the testimony intelligence, not evidence. It can inform public understanding. It can corroborate other findings. It can create pressure for formal investigation. But it cannot, on its own, secure a conviction at the International Criminal Court.

The ICC has mechanisms for protecting witnesses, including anonymous testimony in some circumstances. But those mechanisms are limited. The defence has the right to examine witnesses. The court must be satisfied that the evidence meets the standard of proof beyond reasonable doubt. Anonymous testimony from unknown individuals, without corroboration, rarely meets that standard.

This is the predicament. The most important information is often the hardest to verify. The people who know the truth are often the ones who cannot speak openly. And the systems that produce the harm are designed, in part, to ensure that verifiable evidence does not exist — or that it is buried, destroyed, or classified beyond reach.

III. The Function of Intelligence for the Public

The documentary, and the journalism that surrounds it, serve a different function from a court proceeding. They inform the public. They create a record. They pressure institutions to act. They make the harm visible. They are intelligence for the public.

This is not a lesser function. It is a vital one. Public understanding is a precondition for political action. A population that does not know what is happening cannot demand accountability. A record that is not kept cannot be consulted. A harm that is not made visible cannot be addressed.

But public understanding is not the same as legal proof. And the gap between them is where the systems of harm find their protection.

IV. The Structural Nature of the Predicament

The predicament is not accidental. It is structural.

The systems that produce the harm — AI targeting systems, surveillance platforms, automated decision-making in welfare and disability support — are designed to diffuse responsibility. The machine recommends. The human approves. The chain of command is long. The decision is distributed. No single person feels culpable. And when harm occurs, the question of who is responsible becomes almost impossible to answer.

This is what legal scholars call the “responsibility gap” — a situation where existing law is not equipped to attribute blame to any one human or non-human agent. The machine cannot be held responsible. The human can say the machine recommended it. The system continues.

The gap is not a flaw. It is a feature. It is what allows the system to operate without accountability. If the harm could be traced to a single decision-maker, that decision-maker could be held responsible. But if the harm is distributed across a system — across algorithms, interfaces, protocols, and chains of approval — then no one is responsible. The harm simply happens.

And when testimony does emerge, it often cannot meet the evidentiary standard because the conditions that produce it — anonymity, fear, the absence of documentation — are the same conditions that make it unverifiable in court.

This is the trap. The systems that produce the harm are designed to prevent the evidence that would prove the harm. The people who could provide the evidence are the ones who cannot speak openly. And the institutions that could act on the evidence are the ones that require the evidence to be admissible.

V. What This Means for Accountability

I am not arguing that the ICC should accept anonymous testimony as sufficient for conviction. That would be a different kind of injustice — a system where anyone could be convicted on the word of an unseen accuser.

I am arguing that the gap between intelligence and evidence is a structural feature of the systems that produce harm, and that this gap protects the systems from accountability.

What follows from this?

First, the value of intelligence must be recognised. Journalism, documentary, and whistleblowing serve functions that courts cannot. They inform the public. They create records. They make harm visible. They are not a substitute for legal proof, but they are a precondition for it. Without public understanding, there is no political pressure for formal investigation. Without formal investigation, there is no evidence. Without evidence, there is no accountability.

Second, the evidentiary gap must be addressed. If the systems that produce harm are designed to prevent evidence from existing, then the law must adapt. This means:

· Stronger protections for whistleblowers and witnesses.

· Independent mechanisms for collecting and preserving evidence, even when it cannot be presented in open court.

· A recognition that the absence of evidence is not the same as the absence of harm — and that systems designed to prevent evidence are themselves a form of obstruction.

Third, the public must be educated about the difference. Most people do not distinguish between intelligence and evidence. They see a documentary and assume it proves something. Or they see a court acquittal and assume nothing happened. Neither is correct. The public needs to understand that knowing and proving are different acts, and that the gap between them is where accountability is lost.

VI. The Desert and the Boundaries

I have written elsewhere about the desert — the zones where the state has withdrawn care and left the vulnerable to fend for themselves. The desert is real. It has boundaries. And those boundaries are guarded by mechanisms that prevent the truth from being proven.

The AI targeting system is one such mechanism. It reduces human beings to data points. It diffuses responsibility across a chain of algorithms and approvals. It creates a language — “collateral damage,” “targeted killing,” “threat score” — that obscures the human reality of what is being done.

The surveillance platform is another. It monitors populations. It identifies dissent. It creates a chilling effect that silences witnesses before they speak.

The classification system is another. It buries information beyond the reach of courts. It makes the evidence unavailable. It protects the system from scrutiny.

Together, these mechanisms form the architecture of impunity. They ensure that the harm can be known — by those who experience it, by those who witness it, by those who document it — but not proven. And in the gap between knowing and proving, the systems continue.

VII. A Question for the Reader

The systems that produce harm are designed to prevent the evidence that would prove the harm. This is not a conspiracy. It is the operational logic of power when it has nothing to fear from accountability.

So the question is not whether the harm is happening. The question is whether we will continue to accept the alibi. The alibi of the machine. The alibi of the system. The alibi of the chain of command. The alibi that says: no one is responsible because everyone is responsible.

The question is whether we will insist on accountability even when the evidence is hard to obtain. Whether we will build mechanisms that can trace responsibility through the labyrinth of algorithms and approvals. Whether we will educate the public about the difference between knowing and proving — and demand that the law adapt to the reality of the systems it is supposed to govern.

The desert is real. It has boundaries. And the boundaries are guarded by the mechanisms that prevent the truth from being proven.

We know what is happening. The question is whether we can prove it. And if we cannot prove it, whether we will continue to look away.

A note on sources and method: This essay draws on publicly available journalism, academic research, and the author’s own work on automated decision-making in government. Where specific systems and events are referenced, they are presented as reported, not as established fact. The distinction between intelligence and evidence is central to the essay’s argument, and the same distinction applies to the essay itself. It is written to inform public understanding, not to prove a case. The reader is encouraged to check the sources, test the argument, and reach their own conclusion.

— Andrew Klein