AWS Continuum GA Prices Validation at $10 per Inference Unit

AWS has made Continuum for code vulnerabilities generally available in US East (N. Virginia) and US West (Oregon), four months after a gated preview, InfoQ's Steef-Jan Wiggers reports. The service validates scanner findings, then opens pull requests for fixes. Its pricing page, absent from the AWS launch post, charges by both lines of code protected and per finding worked.

The problem AWS is targeting is volume, not detection. Swami Sivasubramanian, VP of Agentic AI, and Christopher Rae, who leads AI security go-to-market at AWS, both describe large application estates receiving thousands of scanner findings a day, many unreachable, unexploitable, or already mitigated by runtime controls. Continuum's response is a context graph built from code repositories and the surrounding AWS environment, turning raw findings into ranked, validated issues with attack paths and blast-radius assessments.

The service ingests code from a GitHub integration or an S3 bucket in the same account and Region, and can optionally consume a collector placed in the account and Region where the code runs. The collector gathers AWS Config resources and network configuration so Continuum can judge whether a vulnerability is reachable and what an attacker could reach from it. AWS reports each finding receives an evidence level such as Static proven. Where Continuum can reproduce a finding, it runs the exploit in an isolated sandbox and attaches proof-of-concept code with reproduction steps. For confirmed issues, it generates a code patch, dependency upgrade, or configuration change and opens a pull request. In the published console example, a High severity issue with Exploitable status, Static proven evidence, high likelihood, and moderate impact produced a pull request across four files in the cart service, adding 50 lines and removing 24.

AWS's launch post does not mention pricing. InfoQ pulled the model from the pricing page: a context fee covers lines of code protected per month, starting at $200 per million and tiering down to $100 above ten billion lines. An inference fee charges $10 per Continuum inference unit, consumed as findings are worked. AWS's worked examples show the interaction.

In the second example, InfoQ notes, context accounts for $18,200 and inference for the remaining $200,000, or roughly $20 per validated finding.

Only the code vulnerabilities capability reached general availability. InfoQ reports the Continuum user guide sends readers to a separately titled AWS Security Agent User Guide for penetration testing and threat modeling. Penetration testing carries a separate price of $50 per task-hour; threat modeling, announced in preview in June, has no published pricing. AWS serverless hero Yan Cui had told InfoQ in July that customers face "the same capabilities, two product names, and no clarity on which one you should use." That documentation boundary is still visible.

AI Mastery analysis

The per-finding inference fee makes the cost of validation explicit. AWS's worked examples show 10,000 findings at two units each consuming $200,000 in inference, so triage has a marginal price. Bernardo Meireles Correa, an engineer at Avatier, makes the point on LinkedIn: a proof in an isolated sandbox says the code is exploitable, but it does not say the path to it exists in your environment, and findings Continuum cannot prove still need somewhere to go or they quietly become the safe pile. Because the AWS collector is optional setup rather than a default, environmental reachability can be missing from the graph exactly when a team expects validation to have settled it.

This is the same pattern covered in our infrastructure isolation rather than model guardrails analysis: the sandbox proves exploitability of the code, not exposure of the system.

Autonomy raises a second gap. AWS says every remediation currently requires human approval, with future autonomous fixes gated by confidence criteria such as test coverage and historical success rate. Jorg Huser, a former AWS principal, put the acceptance test well: what qualifies a fix class for autonomous merging? InfoQ reports Google's CodeMender already validates patches with an LLM-as-a-judge framework for program equivalence, checking that a fix removes the root cause without altering intended behavior. The AWS console's caveat that a generated patch does not mean the issue is fixed concedes Continuum does not perform that step. Until it does, autonomous merging would automate vulnerability removal without behavioral equivalence evidence.

The signal is broader than one service. AWS has shipped a security workflow with concrete pricing that moves validation from manual triage to paid inference, making the cost of triage explicit. The next frontier is the acceptance criteria for autonomous remediation, because that is where per-finding economics collide with software correctness.

Sources

• Primary: AWS Continuum for code vulnerabilities is now generally available | AWS News Blog
• Reported by: InfoQ AI: AWS Ships Continuum for Code Vulnerabilities