Press Release · Pervaziv AI
Pervaziv AI Launches Cortex AI Repository Defense for Verified Security Repairs
Distributed by DataCult · Originally issued via PRUnderground

Revision-aware security connects source analysis, runtime evidence, AI-assisted remediation and independent validation in one governed workflow.
San Francisco, CA (PRUnderground) November 18th, 2026
Pervaziv AI today announced Cortex AI Repository Defense, a new integrated repository security workflow that connects vulnerability discovery with the evidence needed to evaluate a repair. The release brings source scanning, optional runtime validation, AI-assisted remediation and independent verification into one governed assessment record tied to a specific code revision.
Its goal is straightforward: help engineering and security teams understand what was checked, what changed, what passed and what still requires a decision before software ships.
The announcement marks another step in Pervaziv AI's vision for trusted agentic engineering. Instead of treating security as a list of alerts or a final approval gate, Cortex AI Repository Defense makes the repository revision the reference point for investigation, remediation and release evidence. The result is a practical connection between the people writing software, the people assessing its risks and the people responsible for deploying it.
"An AI system can suggest a fix in seconds. That does not tell a team whether the fix addresses the weakness, preserves expected behavior or even applies to the version they are releasing," said Anoop Jaishankar, Founder and CEO of Pervaziv AI. "Cortex AI Repository Defense connects those decisions to actual evidence. We want security teams and developers working from the same facts, with verification that follows the code rather than disappearing into separate tools."
From Security Alerts to Release Confidence
A security finding is not a verified repair. A passing check may no longer apply to the code being shipped.
The challenge is familiar to organizations shipping software continuously. A scanner identifies a possible access control flaw. A developer prepares a change and opens a pull request. Tests pass. Then a dependency, configuration file or another part of the application changes before release.
The original scan and passing tests remain useful, but neither automatically establishes the security posture of the new release candidate.
Traditional scanning, coding assistance and continuous integration each solve important parts of the problem. Static analysis identifies suspicious source patterns. Coding agents help developers investigate and propose edits. Runtime tests observe application behavior. Pull request checks assess proposed changes.
Yet the handoffs between those activities can leave a critical question unresolved: which evidence supports the software revision that will actually be released?
Repository Defense provides a common assessment record for answering that question. Each campaign identifies the authorized repository revision, requested coverage, progress, findings and supporting evidence. Branches can continue changing while the evidence remains traceable to what was assessed.
If a material input changes, earlier results remain useful history, while the new decision requires applicable current evidence.
From a Plausible Patch to a Verified Repair
AI-assisted code generation can make it easier to draft a change. But a patch is still a proposal.
Security verification asks a different set of questions. Did the change actually address the weakness? Does expected behavior still work? Did the relevant checks pass? And does the resulting evidence belong to the precise revision under review?
For eligible remediation workflows, Cortex AI Repository Defense connects the original finding, affected revision, proposed repair and independent validation results.
Where the project and test environment support it, teams can compare relevant behavior before and after a change. They can examine whether the selected security expectation now holds, whether required regression checks passed and whether the recorded evidence belongs to the pull request candidate being considered.
A repair lacking required evidence remains unverified. An inconclusive check is not converted into success. If a pull request changes, the verification decision must account for that changed revision.
This approach is designed to preserve reviewer judgment while making the evidence behind the decision easier to locate and evaluate.
Repository Defense can also help teams consider findings in context. A severity score is useful, but exposure, potential impact, observed behavior and available exploitation information can all influence priority.
Multiple alerts may stem from one underlying cause, requiring a broader repair rather than a series of isolated edits. AI can assist with interpreting related signals and drafting bounded proposals, while the strength of the security conclusion remains limited by the evidence supporting it.
Start with Source, Then Deepen the Investigation
Cortex AI Repository Defense begins with a quick static examination of an authorized repository.
This gives developers a practical starting point even when a project has no service that can be deployed for testing. Libraries, build tooling, configuration repositories and other software projects can still receive useful source-level review without requiring a runtime environment.
When supported and approved, teams can extend an assessment with deeper source analysis or optional runtime coverage.
Four runtime categories were exercised during initial qualification: scoped discovery of observable application surfaces, passive analysis of web responses, curated HTTP security checks and checks of supported API behavior against selected expectations.
Each category answers a different question, allowing teams to select coverage that fits the project and its risk.
The workflow treats coverage and findings as different measures.
A completed scan with no findings does not mean every possible behavior was tested. Likewise, a runtime category that could not be exercised is recorded as a coverage gap, not silently converted into a passing result.
Even full coverage of four requested categories means those categories were addressed within the selected assessment scope. It does not imply that every possible application behavior has been evaluated.
Evidence That Stays Attached to the Right Revision
One of the defining capabilities of Repository Defense is its treatment of evidence reuse.
Repeating a request for the same authorized project and revision should not always require repeating work that has already produced applicable evidence. Where project, revision, scope and relevant assessment conditions remain equivalent, compatible results can carry forward into a later campaign.
Cortex distinguishes evidence newly executed from evidence reused from earlier authorized work.
That distinction gives reviewers a clearer account of where the result came from and whether a fresh observation is necessary. A source report, for example, may remain applicable when a team expands an assessment to include runtime checks, while the newly requested coverage still requires applicable evidence.
In a November 2026 engineering evaluation, an authorized test project produced two reports for the same source revision.
An initial scoped assessment reported one of one requested runtime categories executed. A later broader assessment reported four of four requested categories covered through eligible evidence reuse across discovery, web, HTTP and API checks. The broader report recorded two findings and six evidence artifacts.
These results demonstrate traceable reuse within that project and assessment scope.
They do not establish a universal acceleration rate, vulnerability detection accuracy or measured improvement in repair success. The practical value illustrated is that reviewers can distinguish newly performed checks, applicable prior evidence and coverage that has not been obtained.
One Security Record Across the Release Decision
Repository Defense is designed to give developers, security reviewers and release owners a common factual record without collapsing their responsibilities into a single automated decision.
A developer needs enough context to understand the finding and prepare a repair.
A security reviewer needs to know what evidence supports the claim, whether relevant behavior was evaluated and what limitations remain.
A release owner needs to understand which checks apply to the candidate being considered and whether required approvals or evidence are still outstanding.
Each role can work from the same repository assessment while permissions and authority remain distinct.
Pull request verification fits naturally into that process. It contributes evidence about a proposed revision. Repository Defense supplies the surrounding context: why the change matters, which finding it addresses and whether the applicable security checks support closing it.
Proposal, verification and authorization remain separate decisions.
Qualified Across Five Cortex Client Surfaces
Pervaziv AI completed focused qualification of the Repository Defense release candidate in November 2026.
The engineering record reports 224 focused core checks and 35 runtime checks passed for the qualification.
Release acceptance covered 19 qualification phases across five client surfaces: VS Code, browser extension, mobile, Cortex CLI and Cortex Discover.
The live release qualification also passed, including acceptance of service readiness, confirmation of the authorized provider revision and fresh authenticated approval evidence.
Together, these checks establish that the examined capabilities met the company's acceptance scope for this release candidate.
Available actions vary by client, permissions and role, and project-specific repair verification remains governed by applicable policy and capability availability.
The broader value of bringing Repository Defense across multiple Cortex surfaces is continuity.
A developer can inspect a finding close to the code. A security reviewer can examine coverage and supporting evidence. A release owner can assess which checks apply to the candidate being considered.
Each role requires different authority, but all can benefit from a consistent account of what happened.
Building on Cortex Code 2.0 and the Cortex AI Model Ensemble
Repository Defense follows the recent introduction of Cortex Code 2.0, which expanded the coding specialist's qualified context range from approximately 128K to approximately 256K tokens.
The long-context qualification produced a 90.3% aggregate pass rate, while the matched standard-task retention suite achieved 100% without a measured regression against its baseline.
These results describe the specified internal evaluations rather than universal outcomes for customer codebases.
That advance gave Cortex more room to examine related source files, requirements, tests, interfaces and configuration during complex engineering tasks.
Repository Defense extends the story beyond understanding and generating code.
It provides the assessment structure in which security observations, candidate changes and eligible verification evidence can be reviewed against the appropriate repository revision.
Within the Cortex AI Model Ensemble, specialized capabilities perform different jobs, including coding, planning, analysis, routing and verification.
Repository Defense is not another specialized model. It is a governed workflow that connects relevant capabilities while preserving independent verification and human authority.
The distinction matters.
More context can strengthen analysis of a complex codebase. A useful patch can accelerate remediation. Neither substitutes for checks establishing what was observed after the software changed.
Pervaziv AI's broader direction is an enterprise AI experience that takes work from intent through execution to evidence rather than stopping at a convincing answer.
From Repository Defense to Collective Cyber Defense
Software supply chains connect maintainers, enterprises, service providers and public institutions.
A project developed by a small team may eventually be deployed in systems and environments that team has never seen. When maintainers can examine an authorized project, prepare a repair with clear evidence and communicate affected versions responsibly, downstream users have a stronger basis for deciding how to respond.
Pervaziv AI joined the Open Secure AI Alliance, founded by NVIDIA and other industry participants and now hosted by the Linux Foundation.
The alliance brings organizations together around open security tools, research and shared defenses for AI systems.
Repository Defense applies similar principles at the software repository level through authorized defensive work, visible assessment scope and evidence that authorized reviewers can examine.
Pervaziv AI is also a signatory to OpenAI's Call for Collective Action on Cyber Defense, an initiative focused on broadening access to AI-powered defensive capabilities and improving coordination across organizations.
Repository Defense brings that commitment into day-to-day engineering by helping maintainers and software teams identify weaknesses earlier, evaluate repairs and provide release owners with clearer evidence for action.
The two efforts address complementary responsibilities.
One focuses on strengthening shared defenses around AI and its supporting infrastructure. The other focuses on making useful defensive capabilities available to more organizations.
At the repository level, that means lowering the barrier to the first assessment while keeping permissions, evidence, remediation authority and responsible disclosure in the hands of the people accountable for the software.
Get Started with Cortex AI Repository Defense
Cortex offers a freemium starting point.
Developers can use the Free Tier for no-cost file scanning or, where eligible, begin a trial to assess an authorized repository, including an open-source project they maintain or have permission to examine.
Teams can review initial source findings before deciding whether deeper analysis, runtime validation or additional organizational controls are appropriate.
Trial availability, repository limits and advanced capabilities depend on the selected plan.
The new release reinforces Pervaziv AI's focus on trusted developer tools, cybersecurity automation and measurable enterprise AI.
Finding a vulnerability is important.
Proposing a fix is useful.
Knowing what was checked, which software revision the evidence applies to and whether the repair has the support required for approval is what turns security work into a defensible release decision.
Product Installation: https://pervaziv.com/start-here/.
Press Contact
Name: Contact Us Email: Contact UsAbout Pervaziv AI
Pervaziv AI develops Cortex, an enterprise AI platform spanning agentic software engineering, cybersecurity, private intelligence and governed workflows across developer and work surfaces. Pervaziv AI focuses on connecting intelligence, execution and evidence across the software lifecycle.