OpenAI and Anthropic Eye Cybersecurity as Next Revenue Driver

The leading AI labs are preparing to monetize a problem they helped create: as their models become increasingly capable of exploiting software vulnerabilities, OpenAI and Anthropic are building cybersecurity offerings into major revenue streams.

The timing is strategic. Both companies need to sustain aggressive growth ahead of anticipated IPOs, and established categories like coding assistance may not be sufficient. OpenAI has reached $40 billion in annual recurring revenue, while Anthropic has reportedly crossed $65 billion, but some major customers have grown more cost-conscious in recent months.

Frontier Models Excel at Security Tasks

Multiple startup founders report that current AI models demonstrate exceptional capability at identifying and patching system vulnerabilities. Erik Bernhardsson, co-founder of infrastructure company Modal, said his team has been using these models as a partial replacement for expensive external security consultants. "Where there's a vulnerability in a system you can now hack something in a few hours," he noted. "We've been running those models internally; they're very good at finding things."

Jean-Denis Greze, co-founder of AI assistant startup Town, said continuous security monitoring has become economically viable for the first time. "Before it wasn't worth it — now it is," he explained.

OpenAI signaled its commitment to the category by holding a dedicated cybersecurity session led by president Greg Brockman during its recent Astra model launch. CEO Sam Altman has actively promoted the cyber capabilities. Nvidia CEO Jensen Huang highlighted AI cybersecurity as the next major application at a Goldman Sachs conference.

The Irony of Selling Solutions to Self-Created Risks

An Anthropic threat intelligence report published this week detailed how bad actors are attempting to use Claude for malicious purposes, underscoring the dual-use nature of these powerful models. The same capabilities that make them effective at defense also make them potent offensive tools.

One open question is whether demand will prove durable. Some founders suggest much of the current opportunity involves patching vulnerabilities in legacy human-coded systems. Once those are addressed, the need for continuous AI-powered security monitoring could decline. However, a broader industry shift is underway, with cybersecurity budgets moving toward AI tools and away from traditional software from companies like Cisco and SentinelOne.

Why it matters

This represents a fundamental shift in how AI labs will generate revenue. Rather than relying solely on token consumption from coding and chatbot applications, they're positioning themselves as essential infrastructure for defending against AI-enabled threats. It's a self-reinforcing business model: as AI models become more capable, they create both the security problems and the solutions enterprises will pay to deploy.

These details were first reported by Newcomer.