OpenAI President: Security Spend Should Vastly Outstrip Software Spend
OpenAI pulled about 25% of its production engineers off their projects to find security holes. Greg Brockman says security spend should vastly outstrip software spend now that code is cheap.
· 7 min read
Greg Brockman, co-founder and president of OpenAI, interviewed by Sarah Guo and Elad Gil (No Priors)
Greg Brockman is a co-founder and the president of OpenAI. He studied math before turning to software, was an early engineer and CTO at Stripe, and helped start OpenAI in 2015. He now oversees its research and product direction, including a consumer base of about 1.2 billion ChatGPT users and a leading coding and enterprise business. In this conversation he covers the company's latest math results, the new proactive agent called Dots, and a billion-dollar program to put frontier models in the hands of cyber defenders.
Summary
Greg Brockman says AI can now generate new knowledge, pointing to almost 400 open math problems that OpenAI's models resolved. He expects biology to follow, with data and regulation as the bottlenecks, and thinks the best interface is no interface, with the AI shaping a custom UI when you need one. On cybersecurity he says the world is in a defender's window while frontier models still favor the people patching, and OpenAI froze 25% of its production engineers to secure its own systems. Because software is becoming nearly free, he argues, security spend should vastly outstrip it. These signals are from an interview that Sarah Guo and Elad Gil did with Greg Brockman on No Priors.
