Why People Wait Until They Leave to Tell the Truth

In The Atlantic this week, David Robinson, who led the writing of OpenAI's safety reports, explained why he resigned. In I Quit OpenAI Because Its Culture Is Broken, he argues that the deeper issue in AI safety is not a missing rule or law. It is culture: perpetual sprints, extraordinary confidence and a habit of fixing problems after they appear. He calls for the kind of layered rigor that nuclear plants and airports depend on.

It is a serious, measured piece. He credits his former colleagues as smart and hardworking, and OpenAI stands by its safety practices. I will leave the technical debate to the experts. What caught my attention is a human detail that applies to every organization, not just AI labs.

Robinson writes that perhaps he should have stayed and fought for fundamental change. But in practice, he and his colleagues were "so busy sprinting" that they seldom had the chance to consider big changes, much less make them. So he found his voice on the way out.

The Truth Usually Leaves With the Person

I have seen this pattern in organizations of every size. The most important concerns are rarely secret. People know where the risks are. They talk about them in hallways and side conversations. They wait for the right meeting that never comes. And too often the clearest description of the problem arrives in an exit interview, a resignation letter or, now, a public essay.

The information required to make the right decisions almost always exists inside the building. The conviction to put that information on the table is what is missing.

That is not because people lack courage. It is because the room has taught them what gets rewarded. In a culture built around speed, raising a fundamental concern feels like slowing the team down. Candor starts to look like disloyalty. So people edit themselves, and the organization loses the very insight it needs most.

The data suggests this is common. In a national survey of more than 1,000 U.S. workers conducted for the law firm Outten & Golden, 33% said fear of consequences would prevent them from reporting issues at work. Among workers ages 18 to 34, it was 41%.

Speed Is a Culture. So Is Care.

Robinson draws a sharp contrast between cultures that fix things after mistakes and cultures designed so that human error never becomes disaster. I would add a simpler distinction. Some cultures make it easy to say the hard thing early. Others make it easier to leave.

Standardization is an asset here. Nuclear plants and airports work because their standards are non-negotiable: checklists, redundancy, the authority of anyone to stop the line. But those standards only work because people are expected to use their judgment and voice inside them. The pilot who questions the captain. The operator who flags an anomaly. The engineer who says not yet. That is personalization in its most practical form: every individual's perspective counts at the moment it matters.

In my work on Leadership in the Age of Personalization, the fifth shift moves organizations from results to methods. The question beneath it is: How do you let them do it? A culture that only measures how fast things ship will never hear why something should not ship yet. As I explored in Fear Is a Leadership Choice, organizations often say they want people who challenge the status quo, then quietly penalize them for doing it.

Conviction Before the Exit

I define conviction as the earned capacity to act on what you believe before the outcome is certain. The hardest test of conviction is not speaking up after you have left. It is speaking up while you still have something to lose. Leaders cannot hand people conviction, but they can stop making it so expensive.

As I wrote in AI Brain Fry Is Really a Judgment Problem, the deeper risk with powerful technology is that people stop thinking before the tool does. The same is true of culture. When pace crowds out reflection, judgment is the first thing to go.

Three Questions for Leaders This Week

Whether you lead an AI lab, a hospital or a sales team, ask three questions.

What would our people tell us in their exit interview that they are not telling us today? Ask it directly, and listen without defending.

When was the last time someone slowed us down for the right reason, and what happened to them? The answer tells everyone whether candor is safe.

Is there time built into our pace for big questions? If every week is a sprint, the important conversations will always be postponed.

Robinson closes with a powerful idea: before organizations can teach machines to treat people well, they need to remember how to do it themselves. I would put it this way. The truth should not have to leave the building to be heard.

Want to explore these ideas further? Learn more about my work on leadership, identity and conviction at www.theglennllopis.com.