Category Pressure Report: Enterprise Verification Primitives Fail Under Infrastructure, Identity, and Agentic Load

This report records structural pressure across enterprise security categories and the assumptions each event places under stress.

In partnership with:

Smallstep – Secures Wi-Fi, VPNs, ZTNA, SaaS and APIs with hardware-bound credentials powered by ACME Device Attestation

LockThreat – AI-powered GRC that replaces legacy tools and unifies compliance, risk, audit and vendor management in one platform

CybersecurityHQ operates as an External Cybersecurity Judgment of Record. Weekly Category Pressure Reports examine structural assumption drift across security domains. These artifacts track where control premises are failing, not where incidents occurred. They inform, but do not themselves constitute, published judgments.

Pressure Level: CRITICAL.

Dell RecoverPoint for Virtual Machines (CVE-2026-22769, CVSS 10.0) disclosed February 18. Hardcoded credentials in a plaintext Apache Tomcat configuration file on a disaster recovery appliance. Unauthenticated remote root persistence. Google Mandiant attributes exploitation to UNC6201, a suspected PRC-nexus cluster, active since at least mid-2024. Attackers deployed BRICKSTORM, replaced it with GRIMBOLT in September 2025, created temporary virtual network interfaces ("Ghost NICs") on ESXi hosts for lateral movement into internal and SaaS environments, then deleted the interfaces. Fewer than a dozen confirmed victims. Full scope unknown. CISA added to KEV catalog.

BeyondTrust Remote Support and Privileged Remote Access (CVE-2026-1731, CVSS 9.9) entered mass exploitation within hours of PoC release on February 10. Pre-authentication remote code execution. No credentials required. Approximately 8,500 on-premise instances exposed. Arctic Wolf confirmed structured intrusion campaigns deploying SimpleHelp RMM for persistence and Active Directory enumeration. CISA added to KEV catalog with 72-hour remediation mandate.

Both products sit on the highest-privilege paths in the environment. Neither carried EDR coverage. Both failed at the authentication boundary.

Constraint Logged: Two infrastructure categories designed to protect and recover the enterprise (privileged access management, virtual machine disaster recovery) were simultaneously exploitable through credential and authentication failures. The appliances that organizations trust to verify and restore trust were themselves unverified.

Pressure Level: HIGH.

DigiCert enforces 199-day maximum TLS certificate validity February 24. Sectigo follows March 8. CA/Browser Forum mandatory 200-day enforcement March 15. Code signing certificates reduced to 460-day maximum March 1. Next reduction: 100-day certificates, March 2027. Terminal state: 47-day certificates with 10-day domain validation reuse, March 2029.

Manual certificate management becomes structurally obsolete in five days.

On agent identity: Cloud Security Alliance survey (February 5) confirms the numbers. 23% of organizations have a formal agent identity strategy. 44% authenticate agents with static API keys. 43% use username/password. 28% can trace agent actions to a human sponsor. 18% of security leaders confident their IAM handles agent identities. Keyfactor: 86% say agents cannot be trusted without unique dynamic identities. 55% say leadership is not taking the threat seriously enough.

Constraint Logged: Certificate infrastructure forces automation at the same moment agent identity governance reveals that the dominant authentication method for autonomous systems is the static shared secret. The identity layer is absorbing two simultaneous structural loads: a forced automation mandate on its existing certificate infrastructure, and a new entity class it was never designed to govern.

Pressure Level: HIGH.

Chrome zero-day (CVE-2026-2441, CVSS 8.8) patched February 13. Use-after-free in the CSS engine. Exploitation confirmed before patch. Separately, 287 Chrome extensions with tens of millions of installs found silently exfiltrating browsing histories. Browser trust is degrading on two axes: the rendering engine and the extension ecosystem.

Microsoft February Patch Tuesday: 54 vulnerabilities, six actively exploited zero-days. Multiple flaws bypass security controls (SmartScreen, OLE mitigations, MSHTML). Attackers targeting the trust verification layer, not the application layer.

Two temporal data points define the governance envelope this week. UNC6201 operated undetected on Dell RecoverPoint infrastructure for at least 18 months. BeyondTrust CVE-2026-1731 moved from AI-enabled variant analysis discovery (January 31) to mass exploitation (February 11) in 11 days. Defenders must simultaneously prepare for multi-year dwell times at the infrastructure layer and sub-two-week exploitation windows at the application layer. The governance gap between these timescales is not closing.

Constraint Logged: AI-enabled vulnerability discovery (Hacktron AI finding CVE-2026-1731 by analyzing a related Ivanti bug pattern) compresses the disclosure-to-exploit window below organizational response capacity. The same week, an 18-month nation-state campaign surfaces on infrastructure appliances invisible to detection tooling. Response architecture must span both timescales. No current governance model does.

Pressure Level: ELEVATED.

Vendor governance products are shipping. Enterprise deployment is not matching. The gap is the risk.

Keyfactor: 85% expect agent identities as common as human/machine identities within five years. Only 28% believe they can prevent a rogue agent from causing damage. Strata Identity/CSA: only 18% confident IAM can manage agent identities. 35% rely on shared service accounts. Cybersecurity Excellence Awards nomination data (February 17): agentic AI governance categories among the fastest-growing. NHI and ISPM nominations showing year-over-year growth.

Constraint Logged: The data is now consistent across multiple independent sources. Organizations are deploying agents with human credentials, static API keys, and shared service accounts while acknowledging these methods are insufficient. Vendor availability is outpacing enterprise adoption capacity. The agent identity problem is being named. It is not being governed.

Across infrastructure appliances, privileged access systems, browser engines, certificate authorities, and AI agent deployments, the same architectural condition recurred.

Verification is implemented as executable logic colocated with the systems it governs. Executable logic is adversarial surface.

Infrastructure appliances store credentials in plaintext configuration files and lack EDR. Privileged access tools execute pre-authentication code paths reachable by any network actor. Browsers execute extension-supplied logic with session-level access to all user data. Agents authenticate using shared secrets indistinguishable from the credentials they are meant to protect. Certificate validity windows compress while automation maturity lags behind enforcement deadlines.

Verification layers are not outside the attack surface. They are the highest-privilege components inside it.

This collapses the operating assumption that security controls exist above the adversarial plane. Dell RecoverPoint was the trust anchor for disaster recovery. BeyondTrust was the trust anchor for privileged access. Chrome extensions operated inside the trust boundary of the browser session. Static API keys are the trust anchor for agent authentication. None of these verified themselves before verifying others.

Structural adaptations such as short-lived cryptographic identity, certificate automation, and hardware attestation reduce colocation and compress exposure windows. They narrow the delegation gap. They do not restore sovereign verification.

This condition is formalized in CHQ-P-2026-006: No Enterprise Trust Primitive Is Self-Verifying. The Position governs interpretation of this report.

Unresolved: If verification is executable, and executable systems are adversarial surfaces, what constitutes a sovereign trust primitive inside the enterprise stack?