Enterprise AI Weekly | EP13 | OpenAI Got Hacked By Researchers Using Claude

A small team of security researchers broke into OpenAI this week. Not with a jailbreak, with a bug in a community forum plugin that led to a single sign on flaw, that led to employee ChatGPT and Codex accounts, that led to Slack, GitHub, and a pull request inside OpenAI's own internal codebase. Under seventy two hours, and an AI model did a real share of the exploit work. They were white hats, they reported it, they collected a bounty of about $6,500. That story got a fraction of the attention of the other headline this week, where Dario Amodei called for an AI development slowdown and Sam Altman and Elon Musk both agreed, even as Anthropic walks toward the public markets and OpenAI rules out an IPO this year. In EP13 of Enterprise AI Weekly I take both stories, and make the case for which one should actually worry you. The real risk in enterprise AI right now is not a model getting too smart. It is a very capable model wired into systems nobody governed properly, with identities and connectors nobody is watching. And it connects to something I keep saying: most companies have barely adopted AI at the basic level, real defensible use cases still sit at maybe two to three percent of workflows, and the work that actually pays off is everything that happens before you put a model to work, the data, the governance, and the measurement. Directional as always. Check my math, and tell me if you see it differently. Chapters 00:00 What actually mattered this week, and what didn't 00:45 The path: a forum bug to OpenAI's internal code 02:40 OpenAI's fortnight: breach, outage, lawsuits, new disclosure process 04:15 The slowdown stunt, and its timing 06:30 The real risk: an unpatched library, not a superintelligence 06:35 The adoption reality: two to three percent 08:00 Enterprise AI isn't about the latest model 09:45 The fix: govern, then measure, tokenomics 10:50 Friday cringe, and the closing question Enterprise AI Weekly is a Friday roundup on the economics of enterprise AI, from Paul Brzozowski, co-founder of Olakai. Your AI is an investment. Let's measure it like one. #EnterpriseAI #AISecurity #OpenAI #Tokenomics #AIGovernance #AIROI