Jacob Coxon’s Resignation: A Warning from Inside the AI Revolution

That warning becomes harder to brush aside because it has found an echo within Anthropic itself. Alignment researcher Evan Hubinger has publicly acknowledged the possibility of AI causing human extinction and has assigned a probability of more than 10 per cent to such an outcome within the next decade. That is his personal assessment, not a scientific certainty. But the very fact that researchers at the frontier are seriously discussing such possibilities should force the rest of the world to pay attention.

This is not an argument against technological progress. Human civilization has never advanced without technology. Every great technological transformation has expanded the boundaries of human possibility. The printing press democratised knowledge. The industrial revolution multiplied production. Computers transformed calculation. The internet transformed communication. Artificial intelligence may transform knowledge and problem-solving on a scale we can barely comprehend.

We should welcome that progress. But progress has never meant that human beings must surrender their agency to their creations. That is the line that cannot be crossed.

The question before us is therefore not whether AI will become more intelligent. It probably will. The question is whether intelligence itself will be allowed to become authority. That distinction is crucial.

ALSO READ: Drugged Punjab: The Cost of AAP’s Broken Promise 

A machine may calculate faster than a human being. It may diagnose a disease more accurately. It may identify patterns invisible to the human eye. It may write software, analyse markets, discover molecules and solve problems that once required armies of experts. But superior performance in a particular task does not confer moral authority.

A machine can tell us what may happen. It cannot, by itself, decide what ought to happen. It can optimise an objective. It cannot be allowed to determine whether the objective is worthy. It can offer an answer. Humanity must retain the right to ask whether that answer should be followed. This is where the present AI race becomes troubling.

The danger is not technology itself. The danger is the combination of enormous technological capability, intense commercial competition and inadequate understanding of what increasingly autonomous systems might ultimately do.

Every major AI company has an incentive to move faster. Every country has an incentive not to fall behind. And every researcher knows that if one laboratory stops, another may continue. That creates a dangerous logic - we must build it because somebody else will.

But some technologies require a different logic. If an AI system becomes capable of improving its own capabilities, influencing other systems and operating with increasing autonomy, the consequences of losing control may not be comparable to an ordinary technological failure. A defective smartphone can be replaced. A failed software update can be reversed. A dangerous algorithm can, in principle, be shut down.


The world now needs to insist on asking equally forcefully - “Should we deploy it, under what conditions, and who remains responsible?” There is nothing anti-scientific about asking those questions. On the contrary, responsible science has always depended upon them.

Nuclear physics did not become dangerous because scientists studied the atom. Biotechnology did not become dangerous because scientists studied genes. The challenge in every technological revolution has been to ensure that human wisdom keeps pace with human capability. AI may be the greatest test yet. Because this time the technology concerns something particularly close to the definition of humanity itself: intelligence.

For the first time, human beings are deliberately constructing systems whose principal purpose is to reproduce, extend and eventually surpass aspects of human cognitive ability. That should inspire wonder. It should also inspire humility. We should not fear a machine simply because it is intelligent. Nor should we worship it because it is intelligent. We must remember who created it and why.

The purpose of technology is to enlarge human freedom, dignity and capability. AI should help the doctor, not make the doctor morally irrelevant. It should assist the teacher, not make education merely an automated transaction. It should help the scientist, not become an authority beyond scientific scrutiny. It should strengthen institutions, not quietly replace human responsibility.

Most importantly, it must not turn human beings into passive consumers of decisions made by systems they no longer understand. That would be a peculiar definition of progress: creating machines to make life easier, only to discover that human beings have become less capable of deciding their own lives. There is also something deeply human in the fear expressed by those who are sounding the alarm.

ALSO READ: Old wounds of Kashmir Valley are being weaponised again 

Humanity has always been fascinated by its own creations. We build monuments, machines and institutions and then become dependent upon them. But AI presents a more intimate possibility. We may create something that does not merely extend our hands or our memory, but our intelligence itself. And if we are careless, the creator could gradually become subordinate to the creation.

That possibility should trouble us  not because it is certain, but because the consequences would be so profound if it were ever to occur. The answer is not to stop innovation. The answer is to put the human being at the centre of innovation.

Human control must not be an optional safety feature. Human accountability cannot be outsourced. The ability to refuse, interrupt, override and ultimately shut down a system must remain with human institutions operating under law and democratic accountability.

There must always be someone who can say - No. There must always be someone who can ask - Why? And there must always be someone who remains responsible for the consequences. That is what makes this debate larger than Anthropic, OpenAI or any other AI company. It is about the relationship between civilization and the extraordinary power it is acquiring.

Coxon’s decision should therefore not be treated as a prophecy of doom. Nor should it be dismissed as the anxiety of one dissident researcher. It should be treated as a warning from inside the machine. Humanity should listen. We should build better AI. We should make it safer. We should use it to discover, heal, educate and create. We should allow technology to take human beings further than they have ever gone.

But we should never confuse going further with surrendering control.

The greatest achievement of artificial intelligence will not be that machines become more like humans. It will be that humans become more capable because of machines without becoming less human themselves. A civilisation cannot call itself advanced if, in the pursuit of creating a superior intelligence, it makes the human being feel inferior, irrelevant or dispensable. The future does not require humanity to defeat artificial intelligence. It requires humanity to remain human while creating it.