OpenAI co-founder and president Greg Brockman said on the No Priors podcast, published on 15 October 2026, that OpenAI took “about 25% of our production engineers” off their projects to work only on securing the company, using its most capable model against its own infrastructure.
He also said “cyber spend, security spend should actually vastly outstrip software spend,” a view he qualified in the same answer: “I may not be right.” Quotes here come from YouTube’s captions of the episode.
Business Pill · RED TEAMING
A one-minute explainer of red teaming: attacking your own system on purpose to find its weaknesses first. It teaches the general idea only and says nothing about any person or company in this story.
The key insight: As we read it, Brockman described security as a recurring cost that scales with each model release rather than a one-time project: OpenAI turned its strongest model on itself until it stopped finding issues, and he said the exercise repeats with every new model.
Describing how OpenAI responds to each new model, Brockman said: “we took about 25% of our production engineers and said, “All of your projects are on hold. You are just securing now.””
He said the team pointed its strongest model at the time at OpenAI’s own systems, found and fixed issues, and “at some point we saturated the issues that this model was capable of finding.” He said the process has to be repeated with each new model.
He did not give a date for the exercise, the number of engineers involved or the issues found, in the captions we read.
OpenAI’s security loop as Greg Brockman described it on the No Priors podcast published 15 October 2026 (YouTube captions): about 25% of production engineers had projects put on hold, the strongest model was turned on OpenAI’s own infrastructure until it saturated the issues it could find, and the process repeats with each new model.
What Was Posted on X
Convo w @gdb cofounder & President of @OpenAI on @NoPriorsPod
No Priors co-host Elad Gil posted the episode on 15 October 2026 with its chapter list, from AI and math and AI oversight to cybersecurity in the AI age. The quotes in this piece come from YouTube’s captions of the episode itself.
OpenAI took 25% of its production engineers off every project and told them: you are just securing now. Their top model hunted in their infrastructure until it saturated (stopped finding new issues).
Will everyone else follow? New physics of cybersecurity
No Priors co-host Sarah Guo summarised the Brockman episode on 16 October 2026: OpenAI took 25% of its production engineers off every project to focus on security, and its top model hunted through its infrastructure until it stopped finding new issues. It is her summary of what Brockman said in the episode quoted in this piece.
The Spending Claim
One of the hosts noted that security has traditionally been, in the host’s words, “low single digits percent of the IT budget”, and asked what it should be now.
Brockman answered that “cyber spend, security spend should actually vastly outstrip software spend.” He added: “I may not be right”, and tied the view to software becoming cheap to produce, saying buyers want software “that is reliable, that is secure, that you can believe in.”
He gave no figure for what that share should be.
The Defender’s Window
Brockman said “the physics of cybersecurity are changing, and we’re in what we call the defender’s window”: the most capable models can help defenders now, while the capabilities available to everyone keep rising with each release.
He pointed to high-profile bank hacks in South Korea that he said were driven by AI models, and said OpenAI gives defenders access to its most capable models through its Daybreak program.
He also said OpenAI has “a billion-dollar commitment that we’ve made to help frontline defenders get access to our models”, without describing its terms in the captions we read.
Figures Greg Brockman stated on the No Priors podcast published 15 October 2026 (YouTube captions): a billion-dollar commitment to help frontline defenders get access to OpenAI’s models and about 400 open math problems published as resolved; his view that security spend should vastly outstrip software spend came with “I may not be right.”
Math and Drug Discovery
On math, Brockman said OpenAI “published about almost 400 very important open problems that are now resolved”, and that the question of whether AI can produce new knowledge is now “just impossible to deny”.
On science, he said “drug discovery is the most clear and obvious area if you get to pick one that you want to fully revolutionize”, while warning that biology is bound by data in a way mathematics is not, and naming regulatory filings as one bottleneck AI can shorten.
The Structural Read
As we read it, the 25% figure is the operational detail behind the defender’s window Brockman described: if the most capable models can find flaws before broadly available ones can, the company with early access has to spend engineering time on itself first.
As we read it, the spending claim is a bet on where value sits once software becomes cheap to produce. Brockman tied it to software becoming ‘so cheap’ and to buyers wanting software ‘that is reliable, that is secure’, and he qualified it with ‘I may not be right’.
As we read it, the interview links three things we have covered this month: AI-driven bank hacks in South Korea, OpenAI’s Daybreak program for defenders, and the AI-solved math results behind the Simons initiative.
Greg Brockman, OpenAI co-founder and president, on No Priors (15 October 2026)
“people have not yet really internalized how much the physics around cybersecurity are going to change”
Three Implications
FOR SECURITY BUDGETS A host put security at ‘low single digits percent of the IT budget’; Brockman said security spend ‘should actually vastly outstrip software spend’, and said he may not be right.
FOR ENGINEERING TEAMS Brockman said OpenAI put the projects of about 25% of its production engineers on hold to secure its systems, and that the process repeats with each new model.
FOR DEFENDERS He said OpenAI has a billion-dollar commitment to help frontline defenders get access to its models, and gives defenders access through its Daybreak program.
The framework’s thesis: “They are not the same curve, and the gap between them — what the platform makes possible versus what the product surfaces — has a name.”
As we read it, Brockman’s defender’s window is that gap applied to security: the models can already find flaws that most organisations have not yet looked for. The framework does not discuss cybersecurity or this interview.
Watch the Cybersecurity Section
Greg Brockman on No Priors, published 15 October 2026, starting at the cybersecurity chapter (26:08).
What Is Not Established
The quotes come from YouTube’s captions of the episode, which may contain transcription errors; we did not have an official transcript.
Brockman did not say when OpenAI paused the projects, how many engineers that was in absolute terms, or what the billion-dollar commitment consists of. OpenAI did not publish separate figures on these points in the material we read.
Business Engineer Framework
The Product Overhang Doctrine
A Business Engineer framework on the gap between what the technology can do and what users actually do with it.
Greg Brockman said OpenAI moved about a quarter of its production engineers onto security work alone and argued, while saying he may not be right, that security spending should vastly outstrip software spending.
The claims are his, made on a podcast, with no figures on cost, timing or results beyond what he said there.
A note on sourcing. We read YouTube’s captions of the No Priors episode with Greg Brockman, published 15 October 2026, in full; every quotation comes from them and may contain transcription errors. The figures are Brockman’s own statements. We did not contact OpenAI or use other outlets’ reports.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
