News
Where does your data go when the AI reads it?
November 5, 2026 · PenDraco
IBM's 2026 breach study puts prompt injection incidents at $5.89M on average and model inversion at $6.07M. The first question a clinic or an insurer asks is older than AI: does our data leave?
The question every regulated buyer asks first
Does our data leave? For years the honest answer was mostly yes. Intelligence meant shipping data to a platform. You weighed the value against the exposure, and you signed the addendum.
The numbers behind the question
IBM's Cost of a Data Breach Report 2026, built with the Ponemon Institute on 602 organizations, is worth reading before the next version of that conversation.
The global average breach cost reached $4.99 million, up 12 percent. In the United States the average was $11.5 million. One in four malicious breaches were AI-enabled, and those cost about $6 million on average.
The AI-specific incident types carry their own price tags. Breaches involving a prompt injection incident averaged $5.89 million. Model inversion incidents averaged $6.07 million. These are line items in a study of real events between March 2025 and February 2026.
There is a constructive half to the same study. Extensive use of security AI and automation saved an average of $1.93 million per breach. The technology is not the threat. Uncontrolled placement of it is.
What inference placement actually changes
When the inference runs inside the customer's network, three things become true at once. Your data does not leave your walls for processing. The audit boundary is a room, not a contract clause. And the cost stops metering per token, which is what makes a fixed-fee sprint possible for a mid-sized budget.
We ask the placement question first on every engagement. Buyers in regulated lines of work deserve that conversation before the demo, with numbers on the table. Get in touch.
Sources
- IBM and Ponemon Institute, Cost of a Data Breach Report 2026. Released 2026-07-29. Averages and AI-incident figures as quoted above.