Anthropic, OpenAI and Google sell defenses to AI-made risks

Anthropic, OpenAI and Google are increasingly selling protection against threats that their own AI models helped bring about. For corporate customers, the shift points in one direction: cybersecurity budgets that are already large are likely to grow further.

A business built on both sides of the problem

The three companies are best known for developing frontier AI models. Those same models have given attackers and defenders new capabilities, and the labs are now positioning security products and services as a response to the risks that follow. In effect, the same vendors that supply the technology are also offering to help clients manage its downsides.

That dynamic creates an unusual commercial loop. A customer adopting AI tools may need additional safeguards, and the AI developers are well placed to sell them, given their access to the underlying models and their understanding of how those systems behave.

What it means for clients

For companies buying these services, the immediate consequence is cost. Defending against AI-enabled threats adds a new layer to security spending rather than replacing existing tools, so budgets are expected to rise even further.

It also raises a question of dependence. Clients that rely on a vendor for both AI capabilities and protection from AI-related risks concentrate more of their exposure with a single supplier, something security and procurement teams will likely weigh carefully.

The move shows how quickly AI security is turning from a research topic into a product category. How far each of the three companies expands its offering, and how buyers respond to paying for defenses against risks tied to the vendors themselves, will shape the market in the months ahead.