The Technical Context
I skipped the marketing copy and went straight to the pricing page, and it’s pretty straightforward: the AWS Security Agent costs $50 per task-hour, bills per second, and offers a 2-month trial with 200 task-hours per month. For those building AI automation around DevSecOps, this isn't just abstract news—it's a new line item in the budget.
AWS defines a "task-hour" as the agent's active work time during a penetration test run. If a test takes 8 task-hours, that's $400. A typical run on a medium-sized web application might take 24 task-hours, costing around $1,200. For a large-scale environment, this figure can easily exceed $2,400.
One thing I like here is the per-second billing instead of coarse blocks of time. It's a proper engineering approach, avoiding the old cloud circus where you pay for a full hour even if the agent only worked for 11 minutes. But there's a flip side: without discipline in launching tests, costs can add up silently.
The trial seems generous. Up to 200 task-hours per month is enough to run tests on a real product multiple times, review reports, and analyze findings and fix recommendations. AWS also specifies that even public preview participants get this trial, which is fair to early adopters.
What This Changes for Business and Automation
The first consequence is simple: penetration testing in AWS is now easier to integrate into the release cycle as a service rather than a rare "big quarterly check." This fits well with AI implementation and the automation of security gates before production.
Second, teams with a clear testing architecture will benefit the most. Those who run tests on everything without prioritization will quickly see unnecessary bills. In this case, the price enforces discipline.
Third, this is an interesting option for companies that need a managed approach within the AWS ecosystem but don't want to assemble a menagerie of external scanners, manual checks, and custom glue scripts.
I wouldn't view this service as "just another security tool." It's more of a building block for a proper process. At Nahornyi AI Lab, we solve these kinds of problems in practice: we connect security, AI integration, and real-world pipelines so that your team doesn't pay for chaos. If your releases are slowed down by manual checks or process gaps, we can analyze your environment and build AI automation without the extra noise.