The Changing Physics of Cybersecurity and Drug Development with OpenAI President Greg Brockman

Greg Brockman argues that OpenAI’s newest math results mark a real turning point: AI is no longer just solving canned tasks, but generating genuinely new knowledge in domains where humans can verify progress. He says the company has now resolved nearly 400 important open problems, and that the same pattern should extend from mathematics into biology, materials science, and eventually drug discovery because the core ingredients are similar: intelligence, data, and better tools for humans.

He draws a sharp distinction between what is possible and what is already operational. Drug discovery, he says, is the clearest near-term frontier, but biology still has major data bottlenecks and regulatory friction, so progress will require better wet labs, simulations, and workflow automation at every step of the pipeline. He expects AI to massively reduce the time and cost of tasks like regulatory filings, and thinks the industry is only at the start of that transition.

Brockman also makes a forceful case that cybersecurity is entering a “defenders window.” As frontier models become more capable, he believes security spending should rise far above normal software spend, because software is getting cheaper while trust, reliability, and resilience become more valuable. He says OpenAI is using its models internally and through programs for defenders to find and fix vulnerabilities, while keeping humans firmly in control.