Like Bulgakov’s Moscow, AI may seem autonomous -  but the real decisions shaping our world remain human. Responsibility cannot be outsourced to code, systems, or institutions.

AI is now part of our daily lives. That’s no longer speculative -  it’s fact.

It helps draft letters, write reports, summarise documents, answer questions, generate images, tutor students and increasingly shape how we access information. Whether we like it or not, it is here.

And yet, if you listen to public debate, you would think we have unleashed something monstrous.

We are warned that AI is dangerous. That it threatens democracy. That it could manipulate minds. That it might destabilise society. But let us look more closely...

When Bulgakov wrote of Moscow’s elites and demons, he wasn’t lamenting the supernatural - he was warning of how ordinary humans, acting through institutions, can create absurd, oppressive systems. Today, AI risks the same trap: we may see the machine as the threat, while the human hands shaping it remain invisible.

But I keep coming back to a simpler question:

Are we blaming the machine for what is really a human problem?

AI does not pass laws. AI does not close bank accounts. AI does not write regulatory codes. AI does not decide who may speak and who must be silent.

Institutions do.

And here lies the uncomfortable truth.

Throughout history, when powerful tools emerge, institutions gather around them. Sometimes wisely. Sometimes not. Printing presses were licensed. Radio waves were controlled. Television networks were regulated. Financial systems became centralised. Every powerful innovation attracts oversight -  and with oversight comes power.

But something else happens too.

Responsibility becomes diffused.

Instead of saying, “We chose this policy,” we begin to hear, “The system flagged you.”
Instead of, “We made this decision,” we hear, “The algorithm determined the risk.”

The language shifts. Human agency fades into the background. An abstract authority takes its place.

In earlier centuries, institutions sometimes invoked divine authority to justify earthly control. I say this as someone who believes in God: belief is not the problem. But history shows that religious institutions -  like all human institutions - have at times used fear to exercise power. And when harm was done, the blame often fell not on the administrators, but on God.

Now we risk a modern variation.

“The AI decided.”

No, it didn’t.

AI generates outputs based on parameters. It reflects data. It follows constraints. It operates within guardrails designed by people. It does not possess motive. It does not pursue ideology. It does not wake up and decide to reward some and punish others.

If AI is ever used to restrict access, deny services, filter speech, or shape narratives, that will not be the will of the machine. It will be the will of those who control access, enforcement, infrastructure, and policy.

And that is where our attention should remain.

The real risk is not evil AI.

The real risk is the quiet combination of two forces: intentional misuse and slow cultural drift.

In The Master and Margarita, Bulgakov reminds us that oppression often hides in plain sight. The system feels abstract, even inevitable, until satire and chaos pull back the curtain. What seems like an autonomous force – the supernatural in his novel, the algorithm today – is ultimately human-made, human-imposed, human-biased. AI, like bureaucracies and institutions before it, will reflect the choices, oversights, and intentions of those who control it.

Intentional misuse is easy to imagine. Any powerful tool can be used to advance political or ideological goals. AI can amplify messaging, automate enforcement, or prioritise certain perspectives. In the wrong hands, it could become an efficient instrument of control.

But slow cultural drift may be even more dangerous.

Not dramatic tyranny. Not jackboots at the door.

Just convenience.

“It’s easier if the system auto-decides.”
“Appeals are expensive.”
“We need consistent standards.”
“For safety, we must tighten the filters.”

Each step can sound reasonable in isolation. Over time, however, discretion shrinks. Human judgement recedes. Automation expands. Accountability becomes harder to trace.

And when people object, they are told: “It’s not us. It’s the algorithm.”

That is when the messenger begins to take the blame for the master.

Some argue the solution is more bureaucracy - perhaps even a global oversight board to regulate AI development. But history does not give us great confidence that larger bureaucracies automatically produce greater wisdom. Institutions are not immune from ideological capture. They are not immune from power imbalances. They are not immune from self-preservation instincts.

More layers of control do not guarantee more freedom.

Yet the opposite extreme - no oversight at all - is equally naïve. AI is powerful. It can scale influence rapidly. It can embed itself into finance, education, communication, and governance. Pretending that such a tool needs no guardrails would be irresponsible.

So where does that leave us?

Perhaps the answer is neither blind trust nor reflexive fear.

Perhaps the safeguard lies not in expanding bureaucracy, but in strengthening accountability. As of February 2026, the UN has just appointed its 40-member Independent International Scientific Panel on AI (established via 2025 resolution), with Guterres framing it as a step toward "science-led governance" and "human control." The first report is slated before the July Global Dialogue on AI Governance. International coordination may sound reassuring. But history reminds us that large bureaucratic structures do not automatically produce wisdom, neutrality, or restraint.

Oversight must not become insulation from accountability. We have seen this far too often before. 

No irreversible decision without human responsibility.
No enforcement without transparency.
No system beyond scrutiny.
No hiding behind the phrase “the algorithm did it.”

AI should advise. Humans should decide. And humans should be answerable.

Most of all, we must resist the temptation to outsource moral judgement to machines. Technology can assist us. It cannot replace conscience.

If AI truly has the potential to transform humanity in positive ways - and I believe it does – then the transformation will depend not on the sophistication of the code, but on the character of the people who wield it.

The machine is not a god.
It is not a tyrant.
It is a tool.

If we allow it to become a shield for unaccountable power, that will be our doing - not its.

The question is not whether AI can think.

The question is whether we will.

Monty