Continuous Integration

What is changed with security shifting left in DevSecops…

— Nesha Zoric

Short answer: DevOps is the practice of merging development and operations so code ships continuously with low friction. DevSecOps extends that by moving security scanning into the pipeline — SAST on PR, SCA on dependency changes, container scanning on image build, DAST on staging, SBOM on release. The team shape barely changes in 2026 — most companies run a platform team that owns the pipeline and security specialists who own the policies, meeting in the middle. The honest difference is one phrase: DevSecOps makes security a pipeline gate instead of a release-day blocker.

The practices of DevOps and DevSecOps are similar in nature. As long as your team has a CI pipeline and deploys to production from main and runs at least some automated tests, you are doing DevOps. Their PRs get scanned for static analysis, merged branches get scanned for dependency vulnerabilities, images get scanned when built, and a critical finding causes a pipeline failure, not a release-day meeting that’s DevSecOps.

The rebranding is essential as the change in practice is real Let’s take a look at what really changes with scans, where they sit in the pipeline, how team shapes have (and haven’t) evolved in 2026, and what’s marketing versus what’s substance with calling it DevSecOps.

Ruby.ci facilitates pipeline step integration of all tools in DevSecOps toolchain. We’ll say it one last time and drop it.

The plain definitions

DevOps refers to a flavour of culture and tooling. A product's design through production has involvement and accountability of both the developers and operators. The pipeline is self-operating. Daily merging of changes to main. The deploy happens without issue. If you read Martin Fowler on continuous integration or our primer on what CI is, you have the shape.

The process is the same, but security is integrated. The main concept is "shift left" which means implementing security on pipelines like fail and fix and moving it away from release-gate activities which block shipment and are often ignored. The toolchain moves from onetime yearly penetration tests and audits for compliance toward scanning on every change.

Three methods to identify the distinction externally.

  • Does a failed SAST or SCA scan fail the pipeline? If yes, DevSecOps. If no, DevOps with scans tacked on.
  • Does a developer see the scan result inside their PR? If yes, DevSecOps. If the security team runs scans in a separate dashboard, DevOps.
  • Is the SBOM generated on every release? If yes, DevSecOps. If the compliance team maintains an SBOM spreadsheet, DevOps.

Refer to the first hyperlink for an official US government definition of SOW and the second hyperlink for a vendor-neutral framing.

What actually moves to the pipeline

The shift from DevOps to DevSecOps introduces four scanning families in the pipeline. Each has a logical position.

Static Application Security Testing runs on PR open. Analyzes source code for known vulnerability signatures (SQL injection, directory traversal, insecure randomness). We use Semgrep, SonarQube, GitHub CodeQL, Brakeman for Rails. Anticipated duration: 1-5 minutes on a medium repository. Anticipated false-positive rate: elevated at first, reduces after tuning.

It is a software composition analysis which runs on PR open and on dependency updates. Looks for known CVEs in the dependency manifest. Dependabot, Snyk, Trivy, Github Advanced Security. This task should take about 30 seconds. Low false-positive rate but high-rate of false-urgency a CVE in a transitive dev dependency does not carry the same weight as a CVE in the user's auth path.

Scan the container when the image is built Verifies the container image for any OS-level threats. Tools used are Trivy, Grype, Snyk, Docker Scout Anticipated run time: 1 to 3 minutes based on image size.

DAST, which stands for dynamic application security testing, is carried out on the staging environment after deployment, before promotion to prod. Examines the active application for issues related to behavior (XSS, broken authentication, revealed secrets). Tools: ZAP from OWASP, Burp Suite CI StackHawk. Anticipated Duration: 10-60 min.

Software Bill of Materials (SBOM) Generated on Release Lists all shipped dependencies and versions. Syft, CycloneDX, and SPDX tools. Expected execution period: seconds. Anyone that sells software to the federal government must use US Executive Order 14028.

A mature DevSecOps setup has the following order of pipeline.

  • PR open → SAST + SCA + unit tests. All three fail the pipeline if critical.
  • Merge to main → full test suite + build + container scan.
  • Deploy to staging → DAST.
  • Promote to prod → SBOM generation + attestation.

For a stricter mapping, refer to NIST SP 800-204D about secure software supply chains.

Team shapes

Disorientation arises at the team shape inquiry with the rebrand. Most companies don’t actually merge the roles of developer, security engineer, and operator into one person, despite the term suggesting they do.

The team shapes focused on in 2026.

Each product team is assigned a particular security engineer. Costly yet provides the fastest feedback loop available. Uncommon among smaller businesses.

The pipeline is owned by the platform team of security specialists. The policies are owned by security specialists (what is considered critical, which CVEs are blocked, which SAST rules fire). The developers notice successes and failures in their PRs. Most companies that have 50-500 engineers take this shape.

Small companies that purchase Snyk or GitHub Advanced Security and turn the defaults on are not receiving DevSecOps. It is operationally fragile despite being technically accurate.

From 2020, recruiters started posting devsecops engineer jobs, which in practice meant platform engineer who can read a CVE. This toolchain isn’t new; the role is. Refer to our CI/CD tools for DevOps post and CI/CD security for toolchain information.

Where SecOps sits

SecOps or security operations is the older term for the team that runs the SIEM, responds to incidents, and owns detection. Although SecOps and DevSecOps are similar they are not the same.

  • DevSecOps lives in the pipeline. The target is "no vulnerability ships."
  • SecOps lives in production. The target is "we notice when something bad happens."

A well-established firm has both. A startup may not have either and should get DevSecOps first: it is cheaper to fix bugs before than to detect them after.

When the rebrand is marketing

Three indicators that the DevSecOps label is not authentic.

  • The security team runs scans in a dashboard outside the developer workflow. If the dev never sees the finding, it is not shifted left.
  • Policy exceptions are the default. If 80% of findings are suppressed with "accepted risk — do not block", the gate is theatre.
  • The pipeline does not fail on a critical finding. Security teams negotiate pipeline failures per-PR with product teams. If that negotiation happens, the pipeline is advisory, not a gate.

Just because organizations face uncertainties with DevSecOps, does not mean they don’t adopt it. The majority of teams begin with “dashboard scans” and move in the direction of “nobody disagrees with pipeline failures”. The truthful response regarding your position is at the top of the maturity curve.

When DevOps without the Sec is enough

Three instances.

  • Side projects and personal apps with no users. The scan toolchain is more work than the attack surface.
  • Internal tools that never touch the public internet. Still worth having SCA on dependencies. SAST and DAST are overkill.
  • Pre-PMF startups with 1–3 engineers. Ship first, add scans when the company starts selling to customers who ask for SOC 2.

As soon as you close your first serious deal with an enterprise customer, you’ll be forced to adopt DevSecOps. Why? Because security questionnaires that enterprise customers send will compel you to adopt these practices within 18 months. Refer to our take on CI/CD security for the practical starter kit.

How Ruby.ci handles the toolchain

We include a separate scan step in our pipeline for each scan family - Semgrep for SAST, Trivy for SCA and container scanning, OWASP ZAP for DAST and Syft for SBOM. The tools like Buildkite, CircleCI, Jenkins, and TeamCity are open-source there is nothing stopping you from running them anywhere. Our value is the pipeline primitives parallel stages and cache keys that survive scanner updates and result annotations in the PR. Check out our build artifact definition and deployment automation for the pipeline shape surrounding this.

For the deployment side, our continuous delivery vs continuous deployment describes how staging and production gates fit in.

Frequently asked

Check FAQs Section

Still stuck

Why does the pipeline fail to block crucial findings, asks the security team? The practical starting point is our self-hosted runners and CI pipelines guide. DevSecOps is not a product you purchase it is a pipeline you construct.