NIST expands DevSecOps resources while agentic-AI work remains in planning

NIST has expanded its DevSecOps resources with SSDF mapping and implementation examples. Human-directed AI is demonstrated; agentic-AI work remains in planning.

Source published: 2026-09-24

Audience / jurisdiction: United States; NIST guidance and implementation research relevant to information-security and software-development practitioners, not a general legal compliance mandate.

New resources for reviewing development security

NIST’s National Cybersecurity Center of Excellence has announced additional DevSecOps resources, including a mapping to the Secure Software Development Framework (SSDF) and a second implementation covering continuous integration and continuous delivery (CI/CD) automation and container deployment. Published on September 24, 2026, the update gives information-security and software-development practitioners new material to examine when reviewing how security is incorporated into development activities.

The announcement also describes lifecycle scenarios, an appendix addressing SSDF tasks and updated discussion of artificial intelligence. The immediate story is the expansion of implementation resources, not the introduction of a new compliance obligation. For management-system teams, the useful question is whether these examples can help make internal development-control reviews more concrete.

Sources: [1]

Human-directed AI is demonstrated; agentic AI is planned

The distinction between available work and planned work matters. According to the announcement, Build 2 demonstrates human-directed generative AI. Build 3 is still being scoped for agentic AI. These are different implementation stages, and the planned agentic-AI work should not be treated as an already delivered demonstration.

The collaborating teams intend to demonstrate identification, authentication and authorization of AI agents within software development. That establishes a direction for the planned work, but it does not establish that the relevant mechanisms have been implemented or validated. Readers assessing the announcement should keep the delivered Build 2 material separate from expectations for Build 3.

A practical way to preserve that distinction would be to maintain two entries in an internal review: resources available for examination now, and emerging topics to monitor. This would help avoid basing an assurance claim or implementation decision on a demonstration that remains in the planning stage.

Sources: [1]

Turn the mapping into a targeted internal review

The SSDF-to-DevSecOps mapping offers a starting point for a focused internal discussion. As an editorial suggestion, a development lead and an information-security practitioner could select one software delivery workflow, review the mapping and identify where their existing evidence supports the activities being examined. The announcement alone is not enough to determine detailed coverage; that requires reading the underlying material.

For example, a team reviewing a container-deployment workflow could use the second implementation as a comparison point, then record which parts are relevant to its own environment and which need further investigation. Useful review outputs could include unanswered questions, evidence to obtain and decisions requiring an accountable owner.

These are suggested review methods, not mandatory steps. The aim is to use the published examples to sharpen questions about the organisation’s own development arrangements, rather than assume that an example implementation automatically establishes adequate controls.

Sources: [1]

Watch the webinar and consultation dates

The announcement identifies two forthcoming opportunities: a webinar on October 28, 2026, and a November 9, 2026 deadline for comments on updated sections. Both dates are future milestones as of October 10. Neither is a legal commencement date or a certification transition deadline.

Teams considering participation could prepare questions that separate current implementation detail from the planned agentic-AI scope. For example, they could ask how the proposed work will demonstrate agent identity and permissions, while checking the available material before commenting on its coverage. Any submission should be grounded in the sections actually reviewed, rather than assumptions drawn from the announcement.

Sources: [1]

Use the update as guidance, not a compliance mandate

These are United States NIST guidance and implementation-research resources, not an ISO revision, a certification scheme or a new legal obligation. The announcement is relevant to practitioners evaluating software-development security and AI use in development, but it does not establish that all businesses must adopt the examples.

For an organisation operating an information-security management system, a measured next step would be to review the available resources, document their relevance and track the planned work separately. No equivalence to ISO/IEC 27001 or ISO/IEC 42001, or resulting certification impact, is established by the supplied evidence.

Sources: [1]

Sources

1. New NIST NCCoE Resources on DevSecOps and October 28 Webinar on Agentic AI

Your experience

Which part of your software delivery workflow would benefit most from a focused review against the new NIST implementation resources?