Daily Cybersecurity Intelligence — 6 September 2026

Executive Briefing

The 6 September cybersecurity cycle produced six defensible developments after full-source timing, duplication, and materiality controls. The most urgent network-security update was the official NVD publication of six MikroTik RouterOS CVEs at approximately 01:47 IST, following CERT Polska’s 5 September warning that attackers had already combined two RouterOS flaws against internet-reachable SSH services.[1] [2] [3] The sources do not identify which two CVEs form the observed chain, so that mapping remains deliberately unresolved.

Management-plane and AI-development systems also received high-priority updates. NVD published N-able N-central CVE-2026-86218, a CVSS 10.0 pre-authentication remote-code-execution flaw, while N-able’s first-party advisory and fixed build were dated 5 September.[7] [8] OpenMAIC released version 1.0.1 with four coordinated advisories, including a critical cloud-metadata SSRF chain under specific unsafe configuration conditions.[9] [10] [11] [12] [13]

Two target-day disclosures require chronology discipline. SAFE Security’s new Langflow research reproduced CVE-2026-9198, but IBM disclosed and fixed the vulnerability in July and last modified its bulletin in August.[4] [5] [6] Wordfence and WPScan records for MemberDash, Frontend Admin, and Ninja Forms likewise began on 4–5 September, while their CVE publications or material updates landed on 6 September.[16] [17] [18] [19] [20] [21] PostgreSQL’s CNA separately published two PostgreSQL Anonymizer vulnerabilities fixed in 3.1.4, each with important privilege or workflow preconditions.[22] [23] [24]

Verified Developments at a Glance

| Priority | Development | Category | Affected surface | Evidence confidence |

|---|---|---|---|---|

| Critical | Six MikroTik RouterOS CVEs entered NVD after an active-exploitation warning | Network infrastructure | Internet-reachable RouterOS SSH, WebFig, bandwidth-test, TLS, and policy controls | High; exact two-CVE attack chain remains unknown |

| Critical | NVD published N-central CVE-2026-86218 | IT management and remote administration | N-able N-central before 2026.3.1.14 | High; active exploitation not stated in opened sources |

| Critical | OpenMAIC 1.0.1 fixed four coordinated flaws | AI application and cloud security | OpenMAIC 1.0.0 and unsafe access-code, environment, redirect, file, or stored-content paths | High |

| Critical | SAFE published a reproducible Langflow CVE-2026-9198 exploit chain | AI development and application security | Langflow OSS 1.0.0–1.10.0, especially network-exposed auto-login deployments | High for technical reproduction; vulnerability and fix are older |

| Critical | WordPress plugin records exposed two account-takeover paths and a shortcode flaw | Web application security | MemberDash ≤1.8.5, Frontend Admin ≤3.29.12, Ninja Forms 3.14.10–3.15.1 | High for records and fixes; no active exploitation established |

| High | PostgreSQL Anonymizer 3.1.4 fixed two privilege-boundary flaws | Database security | PostgreSQL Anonymizer before 3.1.4 | High; exploit prerequisites retained |

1. Six MikroTik RouterOS CVEs Entered NVD After Active Exploitation

NVD published CVE-2026-67276, CVE-2026-67277, CVE-2026-67278, CVE-2026-67279, CVE-2026-67281, and CVE-2026-86060 between 20:17:17Z and 20:17:18Z on 5 September, or approximately 01:47 IST on 6 September.[3] CERT Polska had disclosed the six flaws on 5 September and separately warned that attackers were actively combining two RouterOS vulnerabilities against internet-accessible SSH services, with successful attacks observed since at least 2 September.[1] [2]

The records cover incomplete RSA public-key comparison during SSH authentication, premature authorization of a related bandwidth-test connection, malformed RSA/PKCS#1 v1.5 signature acceptance during certificate validation, an unauthenticated SSH rekey path that permits command requests, unauthenticated WebFig file disclosure, and an SSH-login argument flaw that can alter a trusted policy mask.[2] [3] CERT Polska and NVD list fixed releases 6.49.21, 7.23.4, and 7.24.2.[1] [2]

The observed MikroTrick campaign must not be over-specified. The directly opened CERT pages confirm a two-vulnerability chain and active attacks but do not map that chain to two named CVEs. The Hacker News synthesis did not add primary evidence that resolves the pair, so no guessed chain is presented.

Defensive actions

Upgrade RouterOS to 6.49.21, 7.23.4, 7.24.2, or a later supported release. Remove direct internet exposure from SSH, WinBox, WebFig, and bandwidth-test services; restrict management to dedicated networks or a controlled VPN; rotate administrative and SSH credentials where exposure existed; and review RouterOS configuration, users, scripts, scheduler entries, files, support outputs, and outbound connections for unauthorized change. Because the exact exploited pair is unknown, patch the full affected surface rather than prioritizing only one CVE.

2. N-central CVE-2026-86218 Became an Official Critical Record

NVD published CVE-2026-86218 on 6 September and describes pre-authentication remote code execution in N-able N-central before 2026.3.1.14.[8] The directly rendered N-able advisory is dated 5 September, assigns CVSS 10.0 and CWE-96, and identifies 2026.3.1.14 as the resolution.[7] The qualifying target-day event is therefore the official vulnerability-database publication, not the vendor disclosure or fix date.

N-central occupies a privileged management position across customer environments. A pre-authentication flaw on an exposed management plane can create downstream risk to credentials, remote agents, administrative workflows, and managed systems. Neither opened source stated that CVE-2026-86218 was actively exploited, so the severity score must not be converted into an exploitation claim.

Defensive actions

Upgrade to N-central 2026.3.1.14 or later, confirm the running build rather than relying only on change tickets, and restrict the management interface to approved administrative paths. Review authentication, web, process, service, agent-deployment, and administrator-change logs from the period of exposure. Rotate high-value credentials or tokens reachable from the management server when compromise indicators exist, and validate that remote agents and automation policies have not been altered.

3. OpenMAIC 1.0.1 Closed Four Security Paths

OpenMAIC published version 1.0.1 at 07:09:09Z, or 12:39:09 IST, on 6 September and released four first-party security advisories.[9] The most severe path, GHSA-9m7h-vh2h-rc3w and CVE-2026-86259, combines fail-open access control with environment-gated URL validation so that unauthenticated requests can reach cloud instance metadata when ACCESS_CODE is unset and NODE_ENV is non-production or unset.[12] [14] [15]

The companion advisories cover an unauthenticated path-traversal and arbitrary JSON-file write that can overwrite classroom data, stored cross-site scripting in persisted slide content, and a redirect-based SSRF bypass on bring-your-own-key base URLs.[10] [11] [13] The file-write advisory explicitly says it is not remote code execution. The stored-XSS path requires a viewer to render the saved content, and the two SSRF paths depend on reachable destinations and configuration.

Version 1.0.1 constrains classroom identifiers, sanitizes stored slide HTML, revalidates each redirect hop, removes credentials on cross-origin redirects, and changes relevant unsafe defaults to fail closed.[9] [10] [11] [12] [13]

Defensive actions

Upgrade OpenMAIC to 1.0.1 or later. Set a strong ACCESS_CODE, use production-safe environment settings, deny metadata destinations at the network and application layers, and use instance-metadata protections that require session tokens where available. Review cloud audit logs, temporary-credential use, outbound requests, classroom and slide changes, and unexpected JSON files. Rotate cloud or model-provider credentials when metadata or server-side request access may have occurred.

4. New Research Reproduced the Older Langflow CVE-2026-9198 Chain

SAFE Security published a 6 September lab analysis of CVE-2026-9198 that chains GET /api/v1/auto_login, when enabled, with POST /api/v1/validate/code.[4] The first endpoint can issue a superuser token to a network caller; the second executes Python decorators, default arguments, and annotations during validation. SAFE demonstrated command execution and reflected output in an isolated Docker environment and documented access-log, process, and credential-review opportunities.

The chronology is older than the article. IBM initially published its bulletin on 2 July, identifies Langflow OSS 1.0.0–1.10.0 as affected, assigns CVSS 9.8/CWE-94, and recommends 1.10.1.[5] The CVE record was published 17 July and updated 17 August.[6] SAFE notes that the official 1.10.0 Docker image disables auto-login at build time, while the unauthenticated chain remains possible where operators explicitly enable it; the code-validation issue still requires the fixed release.[4]

Defensive actions

Upgrade Langflow to 1.10.1 or later. Disable auto-login, place remote access behind authenticated controls such as SSO or mutual TLS, and avoid exposing port 7860 directly. Run Langflow as a non-root service account and restrict its access to model, vector-store, database, and cloud credentials. Hunt for external GET /api/v1/auto_login requests, suspicious POST /api/v1/validate/code bodies, unexpected child processes beneath Langflow or Uvicorn, and unexplained file or scheduled-task creation; rotate exposed secrets after suspected compromise.

5. WordPress Records Highlighted Two Account-Takeover Paths and a Shortcode Flaw

The 6 September CVE publication cycle included two critical WordPress account-takeover records. MemberDash CVE-2026-16310 affects releases through 1.8.5 and allows an unauthenticated registration request with an arbitrary user ID to change another user’s password, including an administrator’s. Wordfence lists 1.8.6 as patched.[16] [17] Frontend Admin by DynamiApps CVE-2026-75816 affects releases through 3.29.12 and allows unauthenticated form submissions to replace an arbitrary user’s email, after which WordPress password reset can complete takeover. Wordfence lists 3.29.13 as patched.[18] [19]

Ninja Forms CVE-2026-80437 is materially different. It affects 3.14.10–3.15.1 and allows unauthenticated users to introduce shortcode syntax through IP or referrer merge-tag values that are later rendered. Exploitation requires a site configuration that displays an affected value, and impact is limited to the capabilities of shortcodes registered on that installation. WPScan lists 3.15.2 as fixed and is withholding proof-of-concept details until 16 September.[20] [21]

Wordfence first published the two account-takeover records on 5 September and updated them on 6 September; WPScan first published the Ninja Forms record on 4 September and updated it on 6 September. No reviewed record claims active exploitation or a confirmed victim count.

Defensive actions

Upgrade MemberDash to 1.8.6 or later, Frontend Admin by DynamiApps to 3.29.13 or later, and Ninja Forms to 3.15.2 or later. Review administrator and high-privilege account password changes, email changes, new registrations, password-reset traffic, active sessions, and plugin configuration. Invalidate suspicious sessions and rotate credentials when account records changed unexpectedly. Inventory registered shortcodes and avoid rendering request-derived IP or referrer values until the Ninja Forms fix is confirmed.

6. PostgreSQL Anonymizer 3.1.4 Fixed Two Privilege-Boundary Flaws

The PostgreSQL CNA published CVE-2026-19633 and CVE-2026-19634 on 6 September; both affect PostgreSQL Anonymizer before 3.1.4 and are fixed in 3.1.4 or later.[22] [23] The project’s technical issue says both paths were reproduced against 3.1.3.[24]

CVE-2026-19633, CVSS 8.8, permits an unprivileged role that owns a masked table and can create objects in a schema to place untrusted operators, domain casts, or view subqueries into a masking expression. Static masking can later evaluate that code with elevated privileges; dynamic masking executes as the reading role, so the privilege outcome is mode-dependent.[22] [24]

CVE-2026-19634, CVSS 6.4, concerns unescaped object names in rules-import functions. A malicious JSON document can inject statements when a privileged operator imports it. Because the functions are security invoker, an unprivileged caller alone does not obtain superuser execution; a privileged import workflow is required.[23] [24]

Defensive actions

Upgrade PostgreSQL Anonymizer to 3.1.4 or later. Inventory masked-table ownership, schema-creation rights, trusted schemas, security labels, custom operators, domains, views, and rules-import workflows. Treat exported or shared anonymization rules as code-bearing artifacts: restrict their origin, review changes, and require controlled approval before privileged import. Examine PostgreSQL audit logs and role changes for unexpected masking operations or privilege grants.

Practical Defensive Actions for the Next Shift

| Time horizon | Action | Validation evidence |

|---|---|---|

| Immediate | Patch exposed RouterOS and N-central management planes and reduce administrative service exposure. | Running-version evidence, restricted management ACLs, exposure scans, and configuration review. |

| Same shift | Upgrade OpenMAIC and Langflow; inspect cloud, model-provider, database, and application secrets reachable from those systems. | Deployment receipts, environment review, cloud audit logs, and credential-rotation evidence. |

| 24 hours | Patch the three WordPress plugins and investigate user-email, password, registration, session, and shortcode-related anomalies. | Plugin inventory, fixed-version confirmation, account-diff review, and reset-log analysis. |

| 48 hours | Upgrade PostgreSQL Anonymizer and review masked-table ownership, schema creation, trusted expressions, and rules imports. | Role and ownership inventory, extension version, import approvals, and database audit logs. |

Watchlist

| Item | What remains uncertain | Evidence that would change the assessment |

|---|---|---|

| HPE ClearPass Policy Manager | HPE’s 2 September notice says an advisory will arrive on 9 September Pacific Time but still provides no CVEs, affected versions, fixed builds, or exploitation status.[27] | The full HPE advisory, CVEs, affected and fixed versions, workarounds, and exploitation statement. |

| myLaurel / DireWolf | A tracker discovered an attacker claim on 6 September, but no company, regulator, law-enforcement, or independent source confirmed an incident or scope.[25] [26] | Provider incident notice, regulator filing, forensic report, or law-enforcement confirmation. |

| RouterOS MikroTrick composition | CERT Polska confirms a two-flaw active-exploitation chain but does not identify which two of the six CVEs are combined.[1] [2] | A primary technical update mapping the observed chain to exact CVEs. |

| DSEWiki / OpenAI | Complete agent ownership, internal task instructions, and OpenAI’s formal incident classification remain unresolved from 5 September. | OpenAI postmortem or independently reviewable execution records. |

| ClickFix / EtherHiding | Initial website-compromise route and unique victim count remain unresolved from 5 September. | Original Netskope publication, hosting forensics, victim notices, or coordinated investigation. |

| JetBrains Cadence and IDScan / Nexus | Downstream credential use in Cadence and the provider, intrusion path, and affected-person count in the ID-document investigation remain unresolved. | Vendor postmortems, cloud forensics, regulator notices, or formal law-enforcement findings. |

Methodology and Limitations

Agent 1 checked the bounded baseline across CISA, national cyber agencies, vendor and incident notices, SecurityWeek, BleepingComputer, The Hacker News, original security research, and policy or law-enforcement sources. The failed original-research lane was rebuilt manually. Six developments qualified because an official record, first-party release, or substantive research publication fell inside the 6 September IST window. Each section separates that target-day event from earlier disclosure, exploitation, or patch dates.

The NVD API returned 99 CVEs in the window, but this report is not an exhaustive CVE feed.[3] High numeric severity alone did not justify inclusion where a brief record lacked a directly verified vendor advisory, reliable fix boundary, or material context. Search snippets, sponsored posts, social claims, copied roundups, and undated pages were not evidence. Absence from the bounded source set is not proof that no other cybersecurity event occurred.

The RouterOS chain’s exact two CVEs remain unknown. N-able’s opened sources did not claim exploitation of CVE-2026-86218. OpenMAIC and WordPress impact depends on stated configurations or workflows. SAFE’s Langflow article is new analysis of an older KEV-listed vulnerability. PostgreSQL Anonymizer exploitation requires documented privileges or operator action. The myLaurel claim remains unconfirmed and is not counted as an incident.

SEO and Suggested CYBER/SEC Internal Links

| Reader intent | Recommended destination |

|---|---|

| Track current exploitation, CVE publications, and management-plane risk | /threat-intelligence |

| Harden routers, remote administration, and exposed network services | /topics/network-security |

| Review metadata-service, SSRF, and cloud-credential exposure | /topics/cloud-security |

| Secure Langflow, OpenMAIC, and other AI application workflows | /topics/ai-security |

| Assess WordPress and database application vulnerabilities | /topics/application-security |

| Apply operational hardening and incident-response checklists | /resources |

| Compare current developments with the 2010–2026 historical corpus | /history |

References