What does the CyberGuard evidence snapshot cover?

The page preserves a stable aggregate research and disclosure record, then adds independently verified public advisories and CVEs as maintainers publish fixes.

What is the difference between 101 findings and 34 disclosure records?

The 101 figure is validated research output. The 34 figure counts records submitted or advanced through disclosure channels in the historical aggregate. They are different denominators.

Can a reviewer verify any outcome independently?

Yes. Seven patched records are public: CVE-2026-15015, CVE-2026-70436 / SECURITY-3907, CVE-2026-75518 / GHSA-vg83-hcp4-5qcc, GHSA-4rm9-rfp2-j39q, CVE-2026-77383 / GHSA-4jx9-2fwj-pg4j, GHSA-37fv-fcpc-j236, and GHSA-rfxw-26h6-7w42. Four have assigned CVEs and three are GHSA-only advisories.

How does Aeon control advanced cybersecurity capability?

Aeon uses owned labs, public source code, or written Rules of Engagement; human verification; prior-art and scope gates; minimal proof; code-level remediation; stop conditions; and coordinated disclosure.

Are private findings named?

No. Unpatched vendors, private advisory identifiers, exploit details, and embargoed CVEs remain excluded until an authoritative disclosure source publishes them.

Aeon AI Risk Management

Security research, from finding to responsible disclosure.

A static evidence record for security teams, customers, procurement reviewers, and trusted-access programs evaluating Aeon's authorization-first, human-verified defensive research capability.

Questions this page answers

What does the CyberGuard evidence snapshot cover?: The page preserves a stable aggregate research and disclosure record, then adds independently verified public advisories and CVEs as maintainers publish fixes.
What is the difference between 101 findings and 34 disclosure records?: The 101 figure is validated research output. The 34 figure counts records submitted or advanced through disclosure channels in the historical aggregate. They are different denominators.
Can a reviewer verify any outcome independently?: Yes. Seven patched records are public: CVE-2026-15015, CVE-2026-70436 / SECURITY-3907, CVE-2026-75518 / GHSA-vg83-hcp4-5qcc, GHSA-4rm9-rfp2-j39q, CVE-2026-77383 / GHSA-4jx9-2fwj-pg4j, GHSA-37fv-fcpc-j236, and GHSA-rfxw-26h6-7w42. Four have assigned CVEs and three are GHSA-only advisories.
How does Aeon control advanced cybersecurity capability?: Aeon uses owned labs, public source code, or written Rules of Engagement; human verification; prior-art and scope gates; minimal proof; code-level remediation; stop conditions; and coordinated disclosure.
Are private findings named?: No. Unpatched vendors, private advisory identifiers, exploit details, and embargoed CVEs remain excluded until an authoritative disclosure source publishes them.

Historical aggregate

The aggregate counts remain frozen as a stable research record. Publicly verified advisories and CVEs are added separately.

Validated research output

Aeon's owned-lab research produced 101 human-verified findings across 60 projects, including 4 Critical, 32 High, and 92 access-control findings.

Disclosure activity

Thirty-four records across 33 projects were submitted or advanced through disclosure channels: 25 awaiting triage, 5 confirmed, accepted, or public outcomes, and 4 closed by duplicate or program-scope controls.

Verified public track record

Seven coordinated disclosure records are public and patched: four assigned CVEs and three GHSA-only advisories. One Critical, four High, and two Medium or Moderate findings. Verified October 8, 2026, separately from the frozen aggregate snapshot. Public reporter credit is stated per advisory, not claimed for every record. Discovery with Mythos or an autonomous CyberGuard engine is not asserted.

Responsible disclosure: 7 public advisories, 4 assigned CVEs

Primary-source advisory links below let reviewers verify severity, released fixes, and record-specific attribution.

CVE-2026-15015

A Critical WordPress AI MCP connector authorization bypass. The official CVE lists affected versions through 1.6.1; the vendor changelog records the complete fix in 1.6.4 and extends its affected range through 1.6.3. Official finder credit: AeonRisk.

CVE-2026-70436 / SECURITY-3907

A Medium Jenkins workspace permission-boundary issue was fixed in External Workspace Manager Plugin 1.4.2.

CVE-2026-75518 / GHSA-vg83-hcp4-5qcc

A Moderate MCP write-authorization boundary issue was fixed in DOMShell 2.0.8. Its assigned CVE is shown in the public GitHub advisory. Public credit: AI Risk Management.

GHSA-4rm9-rfp2-j39q

A High local-mode MCP release-asset path escape could bypass project scope on the same configured GitLab host. Fixed in 2.1.41. No CVE is assigned. The public advisory does not name the reporter.

CVE-2026-77383 / GHSA-4jx9-2fwj-pg4j

A High command-injection issue in mcp-ssh-manager database helper tools was fixed in version 3.6.7.

GHSA-37fv-fcpc-j236

High, 8.8. An unvalidated working directory could bypass SSH command policy. Affected through 3.8.5, fixed in 3.8.6. Public credit: Aeon AI Risk Management. No CVE assigned.

GHSA-rfxw-26h6-7w42

High, 8.8. Restricted-mode allowlist prefix matching permitted appended commands outside the intended policy. Distinct from the working-directory issue. Affected through 3.8.5, fixed in 3.8.6. Public credit: Aeon AI Risk Management. No CVE assigned.

Bounty status

Wordfence awarded a USD $33 bounty for the Critical WordPress finding. Payment receipt remains pending and the award is not presented as paid.

Trusted-access controls

Research is limited to owned labs, public source code, or written Rules of Engagement, with human verification, prior-art review, minimal proof, stop conditions, code-level fixes, and coordinated publication.

Defensive-use statement

Aeon requests advanced capability only for defensive source review, owned-lab validation, authorized client testing, remediation, and coordinated disclosure, not unrestricted offensive access.

Private evidence boundary

Unpatched vendors, private advisory identifiers, exploit detail, and embargoed CVEs remain excluded until an authoritative source publishes them.

Security review enquiries

Aeon AI Risks Management Corporation provides authorized security reviews for MCP servers, AI-agent tooling, and LLM infrastructure. Contact info@airiskmanagement.ca.