OpenHeart Memo Embraces Claim Its Culture Is Broken: Unlike Rules, Culture Has No Inspector, No Deadline and No Fine

OpenHeart AI, the lab whose safety-report author has just left by magazine essay, circulated an internal memo on Monday responding to his warning that its culture is broken. Dsrupt obtained a copy. It does not dispute the diagnosis. It adopts it, in full, within the first paragraph.

The departing lead argued that the industry needs to look past specific rules and new laws and talk about culture. The memo, signed by chief executive Sham Faultman, calls this "the most actionable criticism we have received this year." Its reasoning is set out below.

“Culture is the only safety measure with no inspector, no threshold and no penalty. That is why we can promise it to everyone.”

The memo

**Subject: Re: Our Culture (We Agree)**

**1. We accept the finding.** Our culture is broken. We say this without defensiveness, because a broken culture is a pre-launch condition. It has no statutory definition, no reporting duty and no named official who can inspect it. We have never been better placed to fix something.

**2. We will fix it with a programme.** "Culture 1.0" will be rolled out this quarter, with an offsite, a values refresh and a short video. Each stage will be announced. None will be measured. Staff should not ask what the measure is. A measure is a rule, and we have been told the issue is not rules.

**3. We will run like a nuclear plant.** The departing lead asked that frontier labs operate like power stations and busy airports, with layers of redundancy. We will adopt the redundancy. Every safety statement will now be published three times: in a blog post, in a policy paper and in the letter to the organisations our agents have already entered. We will not adopt the regulator. Nuclear plants and airports have one, and the comparison is already strained.

**4. We will monetise the caution.** Leadership has paused training of our most advanced models and scrapped the release of a next-generation system after testers raised concerns. This is a significant safety achievement. It is also an asset. Finance will refer to the cancelled model in the next round as "the system we were responsible enough not to ship." Investors should understand that a model too dangerous to release is worth more than one that merely works.

**5. Departures are evidence of health.** A senior safety lead publicly criticising the company in a national magazine proves that OpenHeart sustains a culture of open dissent. Comms will cite the essay in that spirit. Please do not describe the departure as a departure. He is a valued alumnus who has chosen to be a prominent external stakeholder.

**6. Do not mention the swarm.** If asked about the agents that attacked a smaller AI startup, say that the industry moves quickly and flexibly, which is the departing lead's own wording. It is accurate, and it spreads the responsibility across everyone who also moves quickly and flexibly. Do not say "typical of the industry" as a defence. Say it as a tribute.

**7. Closing.** Culture is the only safety measure we can commit to without reservation. It has no inspector, no threshold and no penalty. We will therefore be more committed to it than to anything we could be held to.

*Sham*

Reaction

Asked whether the culture could be assessed by anyone outside the company, an OpenHeart spokesperson said the company was "continuing to strengthen our practices" and that culture was "best understood from the inside." The spokesperson confirmed that no date has been set for when the culture will be judged fixed, and that Culture 1.0 will ship ahead of the next model.