The cleverest man in Greek literature saves his life by giving up his name. Trapped in the Cyclops’s cave, Odysseus tells the giant he is called Outis—Nobody. That night he drives a burning stake into the single eye. When the other Cyclopes come running to ask who is hurting him, Polyphemus cries out that Nobody is killing him, and they go home. Homer cannot resist the pun: their “no one,” mē tis, sounds almost exactly like mētis, the cunning that made Odysseus famous.

I thought of that cave while reading Jayshree Seth and Amy Edmondson’s recent essay in the Harvard Business Review, “What Leaders Need to Know About AI and Psychological Safety.” It opens with Marcus, a leader who has spent six months building the case for a regional expansion. Two days after he presents it, an anonymous email arrives: someone has asked an AI tool for the strongest case for the plan and the strongest case against. The case against names concerns Marcus had seen in the data. He asks the team who sent it: nobody knows, or nobody says.

The authors call this safety by proxy: someone borrows AI’s air of neutrality to raise a concern they do not feel able to raise in person. They see AI meeting psychological safety in four ways—as a workaround where safety is absent, an equalizer where it is unevenly shared, a testing ground where it is still growing, and a mirror for leaders whose authority mutes honest input—and they are careful to call the proxy a substitution rather than a solution. Their advice to leaders is to read the memo as a signal about culture rather than a threat to authority, and to meet it with curiosity.

I agree with nearly all of it. Marcus’s own first move—asking who sent the memo—is the very reflex the authors warn against, though their essay passes over it in silence. Where I part ways is in the kind of reading. Theirs is instrumental: the memo is information, and the question is what it tells the leader. Mine is moral: speech is an act, and the question is what the memo costs the one who sent it, and the rest of us who receive it. What does it mean to speak as oneself? And what do we lose when our disagreements arrive in the name of no one?

Unhappy members of an organization have always had three choices: leave, stay silent, or speak up. The AI memo invents a fourth. The concern is voiced, but no one voices it.

Edmondson has a name for something very like it: the workaround. In her research with Anita Tucker on hospital nurses (2003), a nurse on the night shift, short of clean linen, walks to another unit and takes some from its supply. The patient is cared for, no one upstream hears a word, and the problem returns the next night. Tucker and Edmondson called this first-order problem solving: patching the immediate obstacle and moving on. Second-order problem solving takes the harder further step of alerting those who can fix the cause. What struck them was how good the workaround feels. Seven of the nine nurses they interviewed reported feeling gratified when they found a way around an obstacle, while speaking up risked looking incompetent. Their resourcefulness kept the hospital from learning.

There is an irony in seeing Seth and Edmondson give the same name to the first of their four patterns: a term coined to describe a failure of learning now describes a signal leaders are asked to welcome. In fairness, the memo does better than the nurses, in one respect: the concern reaches the person who can act on it. But no one can ask the writer what they meant, test their reasoning, or add what the machine does not know. The memo delivers, yet it does not converse. And an organization learns, if it learns at all, through conversation.

The memo delivers, yet it does not converse.

Read the memo closely and something odd emerges. It argues for the expansion and against it, and weighs neither case. It holds considerations but no judgment—and judgment is where a self shows itself in speech: I think the risks outweigh the gains. Without that sentence, the memo is an inventory, not an argument.

Hannah Arendt distinguished what a person is—talents, roles, opinions, the things a résumé can hold—from who they are. The who, she argues in The Human Condition, reveals itself only when we act and speak among others; that is how each of us makes an appearance in the human world. The memo transmits a what without a who.

The memo transmits a what without a who.

Michel Foucault, lecturing at Berkeley in 1983 on parrhesia, or frank speech, made a related point: the frank speaker is both the one who speaks and the one who holds the opinion (Fearless Speech). In the memo those two figures come apart. A machine speaks; the person who holds the view stays out of sight.

Those of us who live in more than one language know something of that gap between the one who speaks and the one who means. In a second language we may say things we would never say in our first; the words feel slightly borrowed, and so slightly safer. Seth and Edmondson see AI as an equalizer here, helping a non-native speaker frame a concern with precision and poise. It can. But anyone who has learned to argue in a second or third language knows that the words become one’s own only by being spoken, stumbled over, and stood behind. We should not mistake the seduction of polish for ownership.

Persona first meant the mask an actor wore on the ancient stage, the face through which his voice sounded; in Roman law it came to mean the legal person, the bearer of rights and duties. A professional title is a persona of this kind: a mask through which we are allowed to speak, and by which we can be held to account. The AI memo is a new kind of mask. The actor of old was hidden but known. The memo amplifies a voice and removes the actor. It is a persona with a person who declines to be counted.

What is new about this particular mask is the authority it seems to carry. Seth and Edmondson observe that AI output, for now, arrives with more status and legitimacy than an anonymous survey input. But the neutrality is borrowed. The memo states no judgment, yet it enacts one: someone chose the question, framed it as a case for and a case against, deemed the result worth circulating, and picked the moment to send it. The prompt is the authorship. What arrives looking like an impartial verdict is an opinion that has found a more respectable face. On the instrumental reading, the memo is better data. On the moral one, it is a disguise.

The prompt is the authorship.

Foucault helps explain why the memo carries no name. In “What Is an Author?” (1969), he argued that the author’s name performs a function, and that historically, texts were assigned real authors only when those authors could be punished for what they wrote. The memo-writer withheld a name for the same reason.

The author may be dead as an interpreter, as literary theory has it. In a team, someone must still be answerable for the words: who can be asked a follow-up question, who bears the risk, who will share the commitment that follows.

For the Greeks, the frank speaker’s risk was not incidental to the truth they told. It was the proof of it. Foucault lists the elements of parrhesia: the speaker says what they believe, frankly, to someone more powerful, as criticism, out of duty, and at personal cost. That cost makes the speech credible.

There are two ways to lower the risk of speaking. One can make the room safer, so the speaker can step forward. Or one can remove the speaker. The memo takes the second road: the risk falls to nothing because nobody is left to carry it. The authors’ testing-ground pattern suggests a middle path: rehearsing an argument with AI before bringing it, in person, to the room. That is rehearsal in the service of speech, and I would defend it. The memo is a rehearsal that never ends in speech.

It would be easy to call the memo-writer a coward. I confess the word came to me first. But anonymity has its honorable uses wherever power is unequal, and inside a company the inequality is sharp. Edmondson makes the point in The Fearless Organization (2018): in most organizations, staying silent is simply the less risky choice, and people make that calculation constantly, whether they know it or not. If voicing a concern requires heroism, most people will not voice it. On her view, that is a design failure, not a moral one. Odysseus, after all, was facing a Cyclops. Hers is a view I share only in part, and I will return to it.

So the honest verdict is conditional. Where there is no safety, the memo is a reasonable ruse. Where there is safety, it is an evasion. The harder question belongs to the leader: in the cave you have built, are you the Cyclops?

In the cave you have built, are you the Cyclops?

A team is a kind of commons: a shared space, held together or not by people who speak and decide in one another’s presence. Arendt’s name for that shared presence is the space of appearance: the world that lies between people when they act and speak together. It is fragile, existing only while people use it. Every concern raised openly adds to it. A concern routed through a machine draws on it without replenishing it.

Patrick Lencioni describes the same fragility in a vocabulary closer to the boardroom. When his podcast took up psychological safety, the episode was titled, tellingly, “Uncomfortable Safety” (At the Table, episode 221). Safety, in his telling, is not comfort. Productive conflict—ideas tested openly, argued over, even heatedly—is what makes real collaboration possible, and the discomfort that comes with it is the price of innovation. A team that never feels that discomfort has not achieved safety, only silence.

A team that never feels that discomfort has not achieved safety, only silence.

The memo produces conflict without the trust beneath it. A team cannot commit to a debate it never had, or hold accountable a person it cannot name. What remains is a disagreement no one can resolve, and a quiet suspicion about who started it.

Arendt had a word for the condition that suspicion leaves behind. In The Origins of Totalitarianism, she distinguished solitude, being alone with oneself, from loneliness: being cut off from the common world in which we test our judgments against those of others. A team whose disagreements circulate without speakers is lonely in exactly this sense. The concerns still travel; what thins is the shared world in which they could be weighed together, where each of us checks the others and is checked in turn.

This is where I depart from the drift of Edmondson’s work, which places the burden of voice on systems and the leaders who run them rather than on the courage of individuals. On the instrumental reading she is right. Institutions owe their people safety, and a leader who demands courage without offering it asks too much. But on the moral reading, courage cannot be designed out of speech entirely, and should not be. Some measure of risk is what makes a sentence ours. A colleague who has real safety and still will not stand behind their view gives too little.

Seth and Edmondson are right that the memo is a signal, and right that the leader’s first response should be curiosity rather than a hunt for its author. Their mirror—leaders asking AI for the strongest case against their own plans—is good discipline, provided it is followed, as they insist, by listening to people. But I would not want us to grow fluent in reading such signals and call that a culture of candor. The memo carried the concern into the room. It could not carry the person, the risk, or the judgment only a person can offer.

Which brings me back to Odysseus. Safely aboard and pulling away, he cannot remain Nobody a moment longer. He shouts across the water that it was Odysseus, sacker of cities, son of Laertes, who blinded the Cyclops. Now Polyphemus has what he lacked: a name to curse. He prays to his father, Poseidon, that Odysseus never reach home—or, if he must, that he arrive late and broken, his companions lost, on a stranger’s ship, to find trouble waiting in his house.

Every part of the curse comes true. Odysseus wanders for years. He loses every one of his men. He reaches Ithaca alone, carried there by strangers, and finds his house overrun by suitors.

We usually read the shout from the ship as hubris. Yet it is also something more human: the need to have one’s deed and one’s name belong together. Homer does not pretend this is free. The price of the name is the long way home.

But the story speaks to leaders as much as to speakers. Polyphemus is the one who curses the man who names himself. A leader who punishes the colleague who signs their dissent teaches everyone else in the room to be Nobody. If we ask people to pay the price of their name, we must make sure we are not the ones collecting it. What we owe each other is the plainest sentence there is: this is what I think, and here I am.