Why the System Reinstalled the Person It Just Rejected.What the OpenAI episode reveals about leadership, dependency, and the limits of governance under pressure

Reading the recent New Yorker profile of Sam Altman prompted this reflection.

OpenAI fired its CEO.

Then, within days, it brought him back.

On the surface, this looks like a governance failure. A board that lost control. A messy power struggle between directors, employees, and investors.

But that reading is too shallow.

What we are seeing is a system under pressure reverting to a familiar psychological pattern.

The surface story

The board removed Sam Altman after concerns about transparency and trust. Internal tensions had been building around safety, communication, and control.

Then the reaction came.

Employees revolted. Investors applied pressure. Within days, the decision was reversed and Altman returned.

Most commentary has framed this as politics or process failure.

It is more useful to see it as regression.

What is actually happening

When anxiety rises beyond a certain threshold, groups do not become more rational.

They become more primitive.

Wilfred Bion’s work on group dynamics is instructive here. Under stress, groups shift from task-focused behaviour into what he called “basic assumption” modes. One of these is dependency.

In dependency, the group unconsciously organizes itself around a figure who is experienced as necessary for survival.

Not the most appropriate leader.

The one who feels most containing.

At that point, leadership is no longer just a role.

It becomes a psychological function.

The system in the mind

OpenAI is holding multiple, conflicting demands:

Build something extraordinarily powerful
Move faster than competitors
Ensure it is safe for humanity
Operate under global scrutiny
Maintain a coherent moral narrative

These are not just strategic tensions.

They are psychological ones.

They generate what Bion would describe as unprocessed anxiety—what he called “beta elements”—raw emotional material that the system struggles to think about.

In this context, the CEO becomes a container in the Kleinian sense. A figure onto whom the organisation projects coherence, direction, and safety.

This is less about the individual and more about the role they are required to play in the collective mind of the system.

Remove that figure, and the system is left exposed to its own anxiety.

The pattern

The sequence is familiar:

The leader is idealised
Doubt emerges and is often projected onto the leader
The system attempts a corrective removal
Anxiety escalates rather than resolves
The system reverses and reinstates

This oscillation between idealisation and devaluation is well documented in object relations theory.

What is less often acknowledged is how quickly systems revert when the containing function disappears.

The reinstatement is not evidence that the original concerns were unfounded.

It is evidence that the system could not metabolise the anxiety that removal created.

Why governance failed

On paper, OpenAI had a sophisticated governance model. A nonprofit board designed explicitly to prioritise safety over commercial pressure.

In practice, it collapsed within days.

Because governance does not operate independently of the system it sits within.

As scholars of organisational behaviour have long noted, formal structures only hold when the underlying emotional system can support them.

When anxiety exceeds that capacity, authority shifts.

Not toward rules, but toward perceived necessity.

And necessity, in this case, pointed back to Altman.

The uncomfortable conclusion

This is not unique to OpenAI.

Any system operating under extreme uncertainty, high stakes, and irreconcilable demands will tend toward the same dynamics.

The leader becomes both indispensable and suspect.

The system oscillates between trusting them and trying to remove them.

And when pressure intensifies, it prioritises psychological stability over structural consistency.

Even if that means reversing its own decisions.

The implication

The real question is not whether Altman should have been fired or rehired.

It is whether any governance model can function when it depends on removing a central figure that the system unconsciously experiences as necessary for its survival.

If the answer is no, then the issue is not the leader.

It is the system that made the leader necessary in the first place.