OpenAI Fires Researchers Researchers After Alleged Information-Sharing - WSJ
Oh my goodness, hold onto your lanyards, because OpenAI has unveiled the most *streamlined* approach to external collaboration imaginable! ✨ Three researchers on the company's safety team — Jasmine Wang, Tomek Korbak and Mikita Balesni — have been let go over 'alleged misconduct' including sharing confidential company information with a third-party AI-safety organization, per The Wall Street Journal. Three safety specialists, now completely unshackled from internal process and free to think about safety at absolutely any hour of the day. What a generous severance package of pure, unstructured *time*!
And what a *beautiful* word 'alleged' is — such a touching commitment to due process, sitting right there in the same sentence as three full names in a national newspaper. The information-sharing element is even more delightful, because OpenAI has spent years publicly enthusing about external scrutiny, red-teaming and third-party evaluation, and here it is demonstrating that enthusiasm in real time, live, and unmistakably. External safety engagement! Simply with a slightly smaller number of engagements than previously advertised ✨ And three brilliant specialists get to carry all that hard-won expertise out into the wider world, where it will surely be snapped up instantly, at full salary, with a signing bonus.
The sourcing is 'people familiar with the matter,' which means an entire thriving community of knowledgeable insiders is engaged, informed, and — crucially — completely unnamed! Somewhere out there, a third-party AI-safety organization is having the most flattering month of its institutional life: it was apparently important enough to receive confidential information from OpenAI's own safety team, an honour most outside auditors can only daydream about between spreadsheets. And think of the audience! Every safety researcher at every lab now has a crystal-clear, beautifully bounded picture of exactly where the edges are, which is honestly the whole point of a safety culture. ✨ What a magnificent month for knowing things quietly!