AWS Continuum Wants to Close the Gap Between AI-Generated Code and AI-Fixed Vulnerabilities
Amazon's new agentic security service promises continuous discovery, triage, and remediation. In practice, it's a bet that the same AI acceleration creating your backlog can also drain it.
The vulnerability backlog problem is not new. What is new is the rate at which AI coding agents are inflating it. Developers ship more code faster; the findings queue grows faster than any triage team can process it. AWS is positioning a new service, Continuum, as the answer.
Continuum is designed to continuously discover, validate, and remediate vulnerabilities across both first-party and third-party code in enterprise environments. According to AWS VP of Security and Observability Chet Kapoor, the service moves findings through the full remediation lifecycle rather than dumping alerts on a dashboard and walking away. For internally owned applications, it can analyze source code, confirm exploitability, generate fix recommendations, and surface those fixes through existing dev workflows — meaning pull requests your team already reviews, not a separate portal nobody checks.
The failure mode here is obvious: automated remediation tools that lack context make bad calls. AWS has a graduated trust model to account for that. Out of the box, Continuum keeps humans in the loop. Once an enterprise decides the system understands its environment and guardrails well enough, they can flip it into what AWS calls "enforce mode" — at which point the service patches code autonomously. That is a significant amount of trust to extend to any system, and Continuum is still in gated preview, so most teams won't be flipping that switch anytime soon.
Some of Continuum's capabilities are recycled from Security Agent, AWS's existing service, specifically penetration testing and code scanning. The genuinely new addition is threat modeling: Continuum can generate threat models from source code or design documents and output them in STRIDE format automatically. That alone would have saved me several tedious meetings in a past life.
Third-party code is the harder surface. Enterprises often have shallow visibility into what libraries are actually executing in production versus what's merely present in a dependency tree. Continuum's promise — surfacing whether a flagged vulnerable library is actually reachable in a running environment — addresses the real signal-to-noise problem in software supply chain risk. The catch: you can't patch code you don't own. Remediation for third-party findings still means version pinning or compensating controls, which requires someone who knows what they're doing to implement correctly.
The CISO role shift Continuum implies is real regardless of whether this particular service delivers. The work moves from managing a findings dashboard to governing the rules under which an automated system acts — what it can touch, what needs sign-off, and what risk level is acceptable in production without human review. Less manual triage, more policy authorship.
Operational takeaway: before you think about enforce mode, make sure your guardrails are documented, tested, and owned by someone whose job it is to update them when the environment changes.
