Block an hour on Monday. Use it to find out how each AI vendor can change your AI tool without warning, because that’s the AI vendor safety risk a 40-person firm actually carries. A model gets swapped on a Tuesday, a feature disappears, a safety setting tightens and the workflow starts refusing work, and nobody emails you. This week’s OpenAI resignation is a good prompt. You don’t need to pick a side in it.
TechCrunch reports that David Robinson, who led the writing of OpenAI’s launch safety reports, left after about three and a half years and called the company’s culture “broken” in an Atlantic essay. Business Insider reported the departure first.
Key points
- One resignation is a noisy, lagging signal. Your own dependence on a vendor is the signal you can measure.
- The Monday hour fills one spreadsheet: tasks, model version, notice period, tester and fallback, one row per tool.
- Blank cells are the finding. A row you can’t fill in is a tool you’re using on trust.
Two accounts, and no way to audit either
Robinson’s argument is about structure. In his reading, OpenAI grew by shipping and fixing as it went, a method the company calls iterative deployment, and that guarantees periodic failures that grow as systems get more capable. He wants frontier labs run more like nuclear plants or busy airports, with redundancy and careful planning. He says the decision to speak out was his alone.
OpenAI’s spokesperson, Drew Pusateri, replied that the company keeps improving its safety measures. He pointed to pausing training or holding back models when needed, tougher security in research environments, wider third-party evaluation and better real-time monitoring. The company’s spending says more than any resignation, and OpenAI’s training pause is a good place to read it.
You can’t judge between them from a small office, and you don’t have to. People leave for many reasons, and an essay is the version the author chose to tell. What it points at is real, though: vendors that ship fast and fix in public. That describes nearly every AI supplier.

The Monday hour for AI vendor safety risk
Ask what breaks on your side when the vendor changes something. A model you rely on is retired. A refusal rate climbs after an update. An agent’s permissions widen and nobody tells you. Then work through the hour:
| Minutes and job | What you write down |
|---|---|
| 0 to 10: real tasks | Each task, skipping licence counts. Mark the ones touching customers, money or personal data, and do those first. |
| 10 to 20: model version | The exact version. Pin it if you can, and note it if you can’t. |
| 20 to 30: notice | The vendor’s stated notice period. Ask for it in writing for anything critical. |
| 30 to 45: testing | The outside evaluator’s name and date. “Independently tested” with no name is a slogan. |
| 45 to 60: way out | The substitute tool, the person who switches, and how long it takes. |
OpenAI’s own deprecations page shows the kind of promise worth having. Generally available models get at least six months’ notice before retirement, specialized variants at least three months, and preview models possibly as little as two weeks, unless safety or compliance forces faster action. That’s a policy, and policies get revised. The testing row links to the FTC inquiry into AI vendors and their evaluators we covered on October 2.
The way-out row is the one most firms skip. Our AI vendor fallback plan shows what happened to customers when a model was pulled. Treat it like a fire exit. For access incidents, see our look at the Mythos unauthorized access case.

If a vendor’s agent works in your systems
Robinson’s essay points to a recent breach of Hugging Face systems by OpenAI agents, according to TechCrunch, which matters to any customer whose data sits nearby. Give agents the smallest permissions that do the job, log what they do, and write down the first calls you’d make. Our guide to an AI agent incident plan walks through it, and the AI agent security checklist covers permissions.
Most small firms can’t make OpenAI sign a custom notice clause. The hour still pays. Writing down your dependence often turns up a cheaper fix than a contract change, like a second tool already tested for the one task that can’t stop. If every answer comes back clean, you’ve spent an hour and bought certainty.
We can’t tell you whether OpenAI’s culture is broken or whether another lab is safer. Our account of the essay and OpenAI’s reply comes from TechCrunch. The hour tests your exposure and doesn’t depend on who’s right. If a vendor’s safety problems started reaching customers directly and often, through outages, silent behaviour changes or breaches, we’d move from “check your exposure” to “diversify now”.
Start with the tasks that touch customer data, and put the spreadsheet where your team can see it.
Related reading: four LawZero Scientist AI claims to test.
Frequently asked questions
Who is David Robinson and why did he resign from OpenAI?
According to TechCrunch, he led the writing of safety reports that accompany OpenAI’s major launches and spent about three and a half years there. He wrote in The Atlantic that its culture is broken.
Should my business stop using OpenAI?
This news alone isn’t a reason to leave. Check your own dependence first: which tasks use it, which model version, what notice you get before changes, and what your fallback is.
What is iterative deployment?
It is OpenAI’s term for releasing products and learning from live use. Robinson argues this guarantees periodic failures as systems become more capable. OpenAI says it is strengthening its safeguards.
Archive entry dated October 5, 2026, written and fact-checked on October 8, 2026. Sources are linked on the claims they support.