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.
Key points
- AWS has launched Continuum, a new service built to discover, investigate, and remediate vulnerabilities across first-party and third-party enterprise code.
- For owned applications, it can validate exploitability and push fix proposals through existing developer workflows such as pull requests.
- Continuum carries over penetration testing and code scanning from Security Agent, and adds automatic STRIDE-format threat modeling as a genuinely new capability.
- An "enforce mode" lets the service patch code autonomously, but that requires deliberate trust-building first; the service is in gated preview.
- The CISO's job shifts from working a findings dashboard to writing the governance rules an automated system lives by.
What is Continuum and what does it actually do?
Continuum is AWS's latest attempt to use the same AI momentum inflating the vulnerability backlog to also shrink it. AWS VP of Security and Observability Chet Kapoor described the service in a blog post as moving findings through the full remediation lifecycle rather than depositing them on a dashboard. For internally owned applications, it validates whether a finding is exploitable, proposes a fix, and routes it through pull requests your team already reviews, not a separate portal nobody checks.
On automated threat modeling: we covered STRIDE tooling when it surfaced earlier this year, and Continuum's ability to produce STRIDE-format models directly from source code or design documents is the most practically useful new addition here.
Should you worry about enforce mode?
The failure mode is obvious: an automated remediation tool without context makes bad calls. AWS's answer is a graduated trust model. By default, humans stay in the loop. Once a team decides the service understands its environment well enough, enforce mode is available and the service patches autonomously. That is a substantial amount of trust to hand to any system, and given that Continuum is gated preview, most teams won't be near that decision for a while.
Some capabilities, specifically penetration testing and code scanning, carry over from Security Agent. The rest is new.
Third-party code is the harder surface
Enterprises routinely have shallow visibility into which libraries are actually executing in production versus which are merely present in a dependency tree. Continuum's pitch on third-party code is that it can tell you whether a flagged library is reachable in a running environment, which is the real signal-to-noise problem in supply chain risk. The catch, as Akshat Tyagi of HFS Research noted, is that you can't patch code you don't own. Third-party remediation still means version pinning or compensating controls, which demands someone who knows what they're doing.
A Checkmarx survey we reported on 10 June found nearly half of production code is now AI-generated and shipped despite known unresolved flaws. Continuum is a direct response to that dynamic.
What this means for security teams
The governance shift Continuum implies is real regardless of whether this service delivers on its promises. The work moves from triaging alerts to authoring policy: what the system can touch, what needs a human sign-off, and what risk level production can absorb without review. Less manual triage, more accountability for the rules themselves.
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 things change.
