What is new: The National Institute of Standards and Technology’s National Cybersecurity Center of Excellence (NCCoE) is hosting a free webinar on October 28, 2026, from 1:00–3:00 p.m. EDT about DevSecOps and the impact of agentic super intelligence (SI). The official event page says registration is limited to 500 participants, the event may be recorded, and audience questions or comments may be captured and made public.
Practical next step: If you are considering AI systems that can take actions—not merely generate text—read the official project materials and verify the registration and recording terms before deciding whether to attend.
What the official sources say
The NCCoE event page describes a project update, newly available resources, and a panel on the role of agentic SI in DevSecOps. It says the NCCoE is collaborating with 14 technology companies to demonstrate secure software-development practices and that the webinar will address an agentic-SI use case within that work.
NIST’s project update says the Software and Agentic AI Identity and Authorization project received more than 600 comments on its concept paper. The associated summary of comments says the future work is focused on practical, standards-based approaches for trustworthy identity, authorization, accountability, and governance. NIST says a draft project description will ask for feedback on proposed scope, use cases, architecture, and standards.
Why identity and authorization matter
The project materials describe a distinction that is easy to lose in ordinary AI conversations: an agent may be able to take actions across tools or services, but the system still needs to establish which agent is acting, who authorized it, what authority it has for this task, and how the action can be audited.
The public summary identifies recurring implementation questions around:
- distinct, verifiable identities for non-human agents;
- stable trust anchors combined with short-lived, narrowly scoped credentials;
- least privilege and task-scoped authorization;
- preserving authorization context when work crosses organizational boundaries; and
- traceability back to the responsible human or organization.
These are source-reported themes from public comments, not a final NIST standard or a universal technical prescription.
What this can mean for human-reviewed implementation
Organizational interpretation: A responsible AI implementation should treat identity and authority as part of the workflow design, not as an afterthought. Before an assistant can retrieve information, draft an action, or call a tool, the organization should be able to name the owner, the permitted task, the source boundary, the human decision gate, the audit trail, and the stop condition.
For DTU’s proposed Human-Centered Knowledge-to-Action Pilot, this suggests a useful low-risk design check: keep source-grounded retrieval and educational navigation separate from admissions, licensure, benefits, legal, medical, financial, employment, or disciplinary decisions. Any action that could affect a person’s rights, money, records, or access remains with an accountable human.
What this webinar does not establish
- It does not create a grant, procurement opportunity, credential, accreditation, or student-aid route.
- It does not establish a partnership, selection, endorsement, or eligibility finding for DTU, LBA, or NABA.
- It does not make the voluntary project materials a federal mandate or a certification.
- It does not authorize an organization to submit private data, claim participation, or speak for NIST.
Anyone considering registration should review the official event page, including its capacity, recording, and public-comment terms. The project page also provides a public route for ongoing feedback and a separate Community of Interest form; those are participation routes to evaluate, not evidence of acceptance or collaboration.
A calm implementation checklist
- Name the accountable human and organization.
- Define the exact task and the minimum necessary data.
- Limit tools, permissions, and time to the task.
- Require source links, uncertainty labels, and escalation when the evidence is incomplete.
- Log corrections, overrides, incidents, and the final human decision.
- Stop or redesign the workflow when the source, privacy, accessibility, security, or review boundary fails.
Explore Di Tran University’s public Humanization and implementation work for source-linked learning and practical governance ideas.
This dated, source-linked article is general educational information. Citation is not endorsement, partnership, certification, or independent confirmation of every third-party assertion. Verify the current official source, event status, and your own facts before acting. This is not individualized legal, cybersecurity, financial, medical, or professional advice; use it at your own discretion and risk.