FR
live

AWS Continuum goes GA and ranks code flaws by how close they sit to production

On November 8, 2026, AWS announced general availability of AWS Continuum for code vulnerabilities: a graph links your repositories to your live infrastructure, reproduces exploitable flaws in sandboxes, and ships a fix as a pull request. Measure your exposure window first — roughly twelve hours before a proof of concept circulates — then pilot the tool on a critical perimeter.

A circuit breaker panel in a data center corridor, every breaker switched off and dark except one with a single amber trip indicator lit.

November 8, 2026. AWS announces general availability of Continuum, its code-vulnerability management service. October 2026. The date of the first announcement, still in preview. Twelve hours. The average window, per AWS, before an exploitable proof of concept circulates after public disclosure. Why it matters: Continuum is not another CVE scanner — it closes the loop between code and production, and its value is measured against your exposure window, not its price.

What Continuum does that scanners don’t

A conventional scanner — Snyk, Semgrep, Trivy — emits a list of CVEs with a score attached. Continuum changes the nature of the response. AWS describes it as building a context graph that connects code repositories directly to an organization’s live AWS environment.

The difference comes down to four mechanisms. First, every finding carries an evidence level: the tool does not merely claim a dependency is vulnerable, it weights the confidence of the detection. Second, Continuum reproduces exploitable issues inside isolated sandboxes, running a proof of concept to confirm a flaw is actually reachable — not merely present in a dependency tree. Third, ranking by proximity: findings are ordered by their distance to business-critical systems, turning a flat list into a triage queue a team under pressure can actually work through. Finally, once validated, Continuum delivers the fix ready to review as a pull request, collapsing a cycle that used to take days of cross-team coordination.

The service launches in two regions: US East (N. Virginia) and US West (Oregon). That is a limit to keep in mind for European deployments: data-residency and sovereignty commitments mean you should verify what transits the service before switching it on for sensitive code.

Twelve hours, the number that justifies everything

AWS’s pitch rests on one concrete figure: exploitable proofs of concept now circulate roughly twelve hours after a vulnerability is disclosed. That is the time a team has between a patch going public and exploitation becoming trivial for an opportunistic attacker.

This compression rewrites the economics of manual triage. A classic process — alert received, assigned, prioritized by a human, then validated — rarely takes under a day. Continuum promises to pull that loop under the twelve-hour mark by automating reproduction and prioritization. For an organization with a surface of several hundred services, the argument is not detection precision: it is speed to a merged fix. For a team shipping continuously, the difference between a twelve-hour loop and a two-day loop is the difference between patching before the attacker moves and racing the attacker to the same commit.

The service belongs to a broader trend ETTAYEB has already documented: isolated sandboxes are becoming the cloud’s unit of trust, as shown by AWS Nitro Sandboxes for EKS. The twist here is that the sandbox no longer isolates a production workload — it replays the attack before it reaches production.

What Continuum does not replace

Let’s be honest about the boundaries: Continuum replaces neither a SCA tool nor a linter. It does not analyze dependency debt cold — it only matters for flaws that are actually reachable in your AWS environment. A vulnerable dependency that is never called will stay invisible to it, where Snyk or Semgrep will flag it anyway. The two approaches are complementary: the scanner guarantees coverage, Continuum guarantees priority.

The second blind spot is region. Launched only in US East and US West, the service cannot be used as-is for workloads subject to European data-residency rules. A team handling health or government data will have to wait, or carefully segregate what transits through the service.

Finally, beware the comfort. A tool that ships fixes as pull requests automates the visible part, but moves responsibility onto the human review of that fix. An auto-generated pull request touching critical code deserves the same review rigor as a human patch — otherwise you have industrialized the speed of application, not the quality of the fix.

The rest of the November salvo

Continuum is the most technical announcement, but AWS unrolled a November salvo that tells the same story: the cloud is no longer competing on compute pricing, but on intelligent automation.

Amazon Quick, unveiled at the “What’s Next with AWS” event, is an agentic work assistant wired to files, calendar, and communications, with a desktop app in preview and free tiers. Amazon Bedrock welcomes GPT‑5.5, Codex, and Managed Agents from OpenAI in limited preview — a tie-up that sharpens the contest over enterprise AI platforms. On the FinOps side, Amazon RDS breaks out backup billing at the individual snapshot level, and AWS Lambda adopts Seekable OCI lazy loading: the image is no longer pulled in full at startup, cutting startup times by up to 70%.

AWS is not alone in this shift. Microsoft has been folding agentic security into Security Copilot, and Google Cloud pushes Gemini-driven operations across its stack. Continuum’s edge is the proof-of-concept reproduction: replaying an exploit in a sandbox yields a verdict — reachable or not — that a severity score alone cannot produce. It is also the part rivals will copy fastest, which is precisely why the lead matters now.

Two items complete the security picture. AWS Security Hub now provides an organization-wide SBOM inventory — a direct response to the Cyber Resilience Act, whose SBOM impact ETTAYEB has covered in depth. And AWS Config adds ten new resource types, widening the surface that can be audited as configuration.

The verdict

If your AWS surface spans several hundred services and your Snyk or Semgrep alerts are drowning in noise, pilot Continuum on a critical perimeter — the one touching sensitive data or authentication. The success criterion is not how many flaws it finds, but how far the gap shrinks between detection and a merged pull request. Measure that gap before and after the pilot: it is the only metric that justifies the tool.

If you sit below that threshold, or your workloads live outside US East and US West, a well-run EPSS + KEV triage still suffices — and costs far less. Continuum’s value scales with the size of the graph it can build: below a certain surface, a competent human reaches the same result in a morning. Wait for availability in your region, then reconsider.

References

The cyber brief, every Tuesday

The flaws that matter and the patches to apply, in a ten-minute read.

No spam. One-click unsubscribe.
read next

On the same topic

An AWS outage degrades S3 and DynamoDB and drags fifteen services down with it

Since 14 November 2026, Amazon S3 and DynamoDB have been degraded, and AWS says it is restoring availability through software mitigations under infrastructure constraints. By the morning of 15 November, fifteen of the sixteen services tracked by third-party observers were affected: isolate your S3 and DynamoDB dependencies and revisit your regional failover runbook.

TSMC confirms a leak of 46,000 employee records that came from a cloud provider, not its own systems

On 8 November 2026, an actor put a 6.9GB dataset said to hold 46,000 TSMC employee records up for auction, and on 15 November the chipmaker confirmed in an SEC filing that the information matched data held by a third-party cloud service provider. Its own systems were not compromised: the leak is a reminder that third-party risk is now the weakest link in the cloud chain.

← Back to the feed

Type at least two characters.

↑ ↓ navigate ↵ open esc dismiss