Featured image of post Price of Insecurity: Why Most Companies Choose To Be Insecure

Price of Insecurity: Why Most Companies Choose To Be Insecure

For most companies the rational business choice has been to embrace revenue growth over security. AI has not yet changed any of these incentives but it might.

Corporate insecurity has persisted for decades for reasons unrelated to technology. For most companies, the rational business choice has been to prioritize revenue growth over security engineering.

CEOs are rewarded for quarterly metrics like revenue, new customers, and product launches, whereas a prevented breach tends to be almost impossible to score. CISOs buy dozens of third-party security products to triage alerts, while requests to fix fundamental issues requiring engineering investment are prioritized below the product roadmap.

AI has not changed these incentives yet.

Watch on YouTube.

Consider the commercial incentives of frontier AI labs. Pursuing IPOs and high valuations, labs emphasize how dangerous their models are to sell defensive subscriptions while lobbying to restrict capable open-weight competitors.

Fear sells models and fear sells news.

During the Hugging Face and OpenCode forensic investigations, Anthropic and OpenAI models refused legitimate forensic work. Hugging Face fortunately could switch to an open-weight GLM-5.2 model on its own infrastructure, while OpenCode had to wait for special approval by OpenAI. Defenders need tools they control, free from mid-incident refusals.

AI-assisted attacks are increasing compromise frequency. This could lead to a change in incentives. Hypothetically speaking, when a company gets breached every two months, incident response may consume so much bandwidth that the company can no longer deliver on its business roadmap. That reality may finally give executives sufficient incentive to fund security engineering.

We already know how to secure our systems. Deploy security invariants: infrastructure properties that eliminate attack surface and are enforced without humans or AI in the loop. Treat AI agents as insider risk.

Background music by Jake Lizzio from Signals Music Studio.

The views expressed on these pages are my own and do not represent the views of anyone else.
Built with Hugo - Theme Stack designed by Jimmy