The new GPT-Red model “can break nearly all models it is pitted against,” according to an OpenAI blog post on Wednesday. OpenAI says it used GPT-Red to find vulnerabilities in GPT-5.6 Sol, a process that made it the company’s “most robust model to prompt injections to date.”
Murati, OpenAI’s former CTO (and, briefly, CEO during Sam Altman’s ouster in 2023), wrote on X that the open-weight model, called “Inkling,” was trained from scratch. Judging by the official blog post, though, it seemed like the company was purposefully setting expectations for Inkling low: “It is not the most performant model available today, closed or open … We trained Inkling for solid capabilities across the board rather than state-of-the-art performance in a single area, to serve as a foundation for the models we will train in the future.”
[Thinking Machines Lab]
OpenAI, on its status website, confirmed that there were “elevated errors” affecting ChatGPT. During the outage, when I tried to access ChatGPT on the web, all that loaded was an error message. The company confirmed impacted services had recovered at 8:39PM ET.
Update: OpenAI says the issues are resolved.
[OpenAI Status]
Apple’s lawsuit said that OpenAI “never responded” to outreach in February. But NBC News reports OpenAI did respond, though communication stopped after one of Apple’s lawyers apparently mixed up OpenAI staffers with the last names Wang and Chang.
Here it is, given to Bloomberg:
While we take these allegations seriously, we’re not aware of any evidence that this complaint has merit. We believe in fair competition and allowing people the freedom to work wherever they choose, and we’re focused on building innovative technology that empowers people everywhere.
Apple sued OpenAI on Friday, alleging the company stole hardware secrets.
OpenAI and the prediction market have a deal, The New York Times reports, that pulls in Kalshi data for World Cup-related queries. A prompt for “France and Spain,” for example, included a win forecast attributed to Kalshi.
Prediction markets like Kalshi and Polymarket are everywhere right now, and that’s by design. The companies have inked deals with news outlets, newsletter writers, and random X accounts — and now Kalshi will reach potential customers via chatbots.
And yet, Apple’s accusations around stealing trade secrets might take years to litigate.
The 41-page complaint details how OpenAI allegedly ‘coached’ Apple employees on how to avoid security checks and asked for ‘show and tell’ during job interviews.
Brockman had taken over product when Fidji Simo went on medical leave, and he’ll continue those responsibilities now that Simo is stepping down from her full-time OpenAI job and becoming a “part-time advisor,” CNBC reports.
Thibault Sottiaux, the engineering lead on Codex, clarified on X that the company isn’t sunsetting its AI coding tool despite the launch of ChatGPT Work, a platform that combines both ChatGPT and Codex. Sottiaux teased some other updates coming to the Work app as well, like moving chats and projects into the sidebar.
[@thsottiaux on X (formerly Twitter)]
OpenAI is shutting down its ChatGPT web browser, and there’s only one appropriate reaction.
better-endings:
Atlas? shrug
Get the day’s best comment and more in my free newsletter, The Verge Daily.
