Research about cognitive threats is easy to collect and hard to use. A team can read the papers, track the incidents, and still have no line connecting what it learned to what it changed. Cognitive Sovereignty Framework v3.0, released November 8, 2026, exists to close that gap. Every assessed item now carries a CTSS score, the reasons behind it, source links, an action, and a trigger for further research. Findings connect directly to controls and readiness checks across all six domains. That is the verdict: v3.0 turns a research collection into a working instrument.
The problem v3.0 was built to solve
The human layer has had doctrine since CSF v2.0, released March 30, 2026. Six domains, a threat taxonomy, and the Cognitive Threat Severity Score gave defenders a shared language for how people, not just systems, get attacked. What the doctrine did not give them was a reliable step from a research finding to a control decision. A new paper, incident, or technique could be read, filed, and forgotten, because the link between finding and response lived in someone's judgment and nowhere else.
That gap matters because the human layer is where governance actually fails. A technically sound system can still be operated by a person whose attention, judgment, or moral authorship has been eroded. CSF v3.0 answers by making the connection explicit. The foundation stays intact: six domains, five causal layers, and the existing CTSS formula. What changes is that the research, the scores, the controls, and the assessment workflow are now one connected path instead of four separate artifacts.
One limit belongs up front, because the release body states it plainly. CTSS scores are severity judgments, not measurements. Readers can inspect the evidence, challenge an input, and follow the correction. A score tells you how the assessors judged an item against the evidence they had. It does not measure the threat the way a sensor measures temperature.
What changed: 44 controls become 56
The mitigation catalog grows from 44 to 56 controls. All 44 existing IDs are retained, 20 implementations and readiness anchors are refined, and 12 research-linked controls are added, two per domain. Each added or refined control connects to research and a stated rationale, and the catalog records the action, expected outcome, check, owner role, and response to failure. Prior wording is preserved for review.
| Domain | Controls | Added focus |
|---|---|---|
| Biological | 6 to 8 | Decision pacing and limits on sensitive-state targeting |
| Cognitive | 8 to 10 | Evidence completeness and tracing influence to outcomes |
| Emotional | 6 to 8 | Relationship boundaries, exit, and vulnerability-sensitive advice |
| Social | 7 to 9 | Consent for synthetic testimony and correction across copies |
| Purpose and Moral | 8 to 10 | Visible objectives, conflicts, and challenges to scores |
| Digital and AI Symbiosis | 9 to 11 | Specific authorization, memory, bystander consent, and exit |
The twelve add pacing, consent, exit, and the ability to challenge a score, not new surveillance or enforcement.
Research that stays usable
The release carries 87 individually assessed items from seven weekly collections, preserving original analysis and source context, and 217 tracked source excerpts, including actions, research questions, indicators, watch items, and grouping labels. Individual CTSS inputs and reasons, dated revisions, weekly reports, downloadable data, and recurring-signal flags are included. Protective findings and contrary evidence are retained alongside threat findings.
Two cautions from the release body belong in any honest use of these numbers. Cohort averages describe the items assessed in that collection, and nothing more. Repeated coverage of the same finding does not become independent corroboration. A threat that appears in seven weekly collections is a threat that was covered seven times, not seven independent confirmations.
A workflow you can run
The resources now connect the full process: scope the workflow, score its threats, select controls, test capability, record the decision, assign repairs, and reassess. A reproducible fictional unit example demonstrates the process across all six domains. New readiness-extension checks remain separate from inherited capability totals, so a new check cannot inflate an old score. Missing evidence stays unknown, and synthetic examples stay outside observed results. The framework refuses to let a gap look like a pass.
The workflow ends in a decision record and a repair assignment, not a report.
Easier to explore, and corrected where it was wrong
The Learning Hub, Threat Explorer, Command Center, research hub, and assessment guides are updated, with shared navigation, a new shield logo, a readable README, and searchable research-linked controls. The release declares a U.S. founding-principles baseline with optional perspective templates that preserve the baseline result, so a different viewpoint can be applied without silently changing the score. A data catalog, schemas, agent guidance, bounded read-only queries, and an optional local MCP adapter make the material reusable by machines as well as people.
Seventeen inconsistent reference-score totals were corrected using the original inputs and weights. Calculators and dashboards share generated scores, and one framework-version file keeps current website labels, citation metadata, the registry, and the catalog aligned. Fifty-one automated tests pass, and both GitHub validation jobs passed for the merged update, covering scoring, evidence integrity, revisions, readiness calculations, synthetic exclusion, agent access, and version consistency. One check is still open: rendered desktop and mobile review of the v3.0 materials remains unverified.
Existing threat and mitigation IDs remain usable.
What v3.0 does not claim
The expanded catalog documents what to implement and test. Field effectiveness and observed unit readiness remain unestablished, and the release body says so directly. Neuro-swarm and quantum research remain open, with other roadmap items explicitly marked partial or planned. v3.0 is a stronger instrument, not a field result.
The lineage: v2.0, v2.1, CP.11, v3.0
The version history is the argument for the framework. CSF v2.0, released March 30, 2026, gave the architecture: six domains, CTSS scoring, and a 41-entry Cognitive Threat Taxonomy. CSF v2.1, released July 13, 2026, made the six domains operational, with a readiness rubric for each domain and a mitigation catalog convention that gives every mitigation a persistent ID. The same day, Federal Procurement Integration CP.11 carried eight CSF controls into AI SAFE2 as UAS-H1 through UAS-H8, making the human layer measurable in a federal compliance standard. CSF v3.0, released November 8, 2026, connects the research behind those controls to the controls themselves, and the controls to a workflow a unit can run.
The live /cognitive-sovereignty/ page still states v2.0, with seven mentions and no mention of v2.1 or v3.0. This article is the announcement. The page update follows.
What to do with it
Start with the release itself and the fictional unit example, which shows the full workflow across all six domains. Run one domain through the readiness worksheet before trying all six. CSF defends the human; AI SAFE2 governs the machine. v3.0 is the release where the human side gets an instrument as practical as the machine side already had.
Sources
- Cognitive Sovereignty Framework v3.0 release, GitHub, published November 8, 2026: https://github.com/CyberStrategyInstitute/cognitive-sovereignty/releases/tag/Cognitive_Sovereignty_Framework-v3.0
- Cognitive Sovereignty repository: https://github.com/CyberStrategyInstitute/cognitive-sovereignty
- CSF Learning Hub: https://cyberstrategyinstitute.github.io/cognitive-sovereignty/
- Merged update, pull request 3: https://github.com/CyberStrategyInstitute/cognitive-sovereignty/pull/3
- Release notes audit, November 17, 2026: https://cyberstrategyinstitute.com/ (audit file: ~/workspace/seo-review/release-notes-audit-2026-11-17.md)