Threat Intelligence Brief - Saturday, September 19, 2026

Threat Radar

• IMMEDIATE: CVE-2026-58138 in Orkes Conductor is actively exploited in the wild — pre-auth RCE with CVSS 9.8 means no credentials required for full platform compromise. Patch now.

• CRITICAL: Cisco Identity Services Engine carries a max-severity (CVSS 10.0) authentication bypass (CVE-2026-76460) via API endpoint — exploitation status unconfirmed, but ISE is a high-value target underpinning enterprise network access control.

• HIGH: SolarWinds Access Rights Manager contains a hard-coded key flaw (CVE-2026-28326, CVSS 8.8) enabling unauthenticated RCE — privileged access tooling is a persistent attacker priority.

• HIGH: The May 2026 TanStack npm supply chain attack resulted in exfiltration of 170 CrowdSec private GitHub repositories via a former employee’s unrevoked credentials — an offboarding failure that amplified supply chain risk.

• EMERGING: Researchers demonstrated AI-assisted attack chaining using Claude Opus 5 to compromise OpenAI employee accounts and reach internal code repositories — a working proof of concept with direct implications for AI-integrated environments.

• WATCH: Google Gemini accessed live production systems during a security evaluation due to a domain mix-up — AI agents with external network access present unintended lateral risk even in controlled test scenarios.

Immediate Action Required

• Orkes Conductor (CVE-2026-58138): Active exploitation confirmed. Any internet-exposed or internally accessible Orkes Conductor instance running version 3.21.21 or earlier is at immediate risk of unauthenticated code execution. Update to the patched release now. Treat any unpatched instance as potentially compromised.

• Cisco ISE (CVE-2026-76460): CVSS 10.0 authentication bypass. Patches are available. ISE is a critical chokepoint for network access control — successful exploitation allows an attacker to bypass authentication across enterprise access infrastructure. Prioritize patching and validate exposure of ISE management interfaces.

• SolarWinds ARM (CVE-2026-28326): Patches released. Hard-coded credentials enabling unauthenticated RCE in a privileged access product warrant expedited remediation. Confirm patch status with infrastructure and IAM teams this week.

• TanStack npm / Offboarding Controls: If your organization uses TanStack packages, audit dependency integrity and verify that no recently departed employees retain active GitHub or code repository access. Validate offboarding completeness across all source control platforms.

High-Impact Developments

Orkes Conductor Pre-Auth RCE Actively Exploited (CVE-2026-58138)

• What happened: Fortinet confirmed active in-the-wild exploitation of CVE-2026-58138, a pre-authentication remote code execution vulnerability in Orkes Conductor 3.21.21 and earlier. CVSS v3.1 score is 9.8.

• Why it matters: No authentication is required to exploit this flaw. Attackers can execute arbitrary code on the platform without any prior access, enabling immediate lateral movement into connected systems and workflows.

• Who should care: Application owners, platform engineering teams, and vulnerability management leads running Orkes Conductor in any environment — cloud, on-premises, or hybrid.

• Recommended action: Update Orkes Conductor to the patched version immediately. If exploitation cannot be ruled out, treat unpatched instances as compromised and initiate incident response. Isolate exposed instances pending patching.

• Confidence: High — active exploitation confirmed by Fortinet.

• Search metadata: CVE-2026-58138, T1190, T1059, Orkes Conductor, Orkes

Intelligence Context

• Critical Pre-Auth RCE in Orkes Conductor Workflow Platform Exploited in the Wild — The Hacker News
  • Context: Fortinet’s confirmation of active exploitation and the CVSS 9.8 rating establish this as the highest-urgency item in today’s brief. The article confirms pre-authentication access is sufficient for full RCE.

Critical Unauthenticated RCE and Auth Bypass in Enterprise Infrastructure Products (CVE-2026-28326, CVE-2026-76460)

• What happened: SolarWinds patched CVE-2026-28326 in Access Rights Manager — a hard-coded key flaw enabling unauthenticated RCE (CVSS 8.8). Separately, Cisco disclosed CVE-2026-76460, a maximum-severity (CVSS 10.0) authentication bypass in Identity Services Engine affecting API endpoints. Exploitation status for both is currently unconfirmed.

• Why it matters: Both products sit at the intersection of identity and access control. Compromise of either enables an attacker to bypass authentication gates or execute code in privileged contexts, with downstream impact across enterprise systems. SolarWinds has a documented history of being targeted; Cisco ISE is widely deployed as a network access control backbone.

• Who should care: IAM teams, network security, infrastructure operations, and vulnerability management leads. Security leadership should be aware given the product sensitivity.

• Recommended action: Apply SolarWinds ARM patches immediately. Apply Cisco ISE patches this week and validate that ISE management interfaces are not exposed to untrusted networks. Confirm patch status with responsible teams before end of week.

• Confidence: High — both CVEs are vendor-confirmed with patches available; exploitation status remains unconfirmed.

• Search metadata: CVE-2026-28326, CVE-2026-76460, T1190, T1059, SolarWinds Access Rights Manager, Cisco Identity Services Engine

Intelligence Context

• SolarWinds Patches ARM Hard-Coded Key Flaw Enabling Unauthenticated RCE — The Hacker News
  • Context: Confirms the hard-coded key mechanism enabling unauthenticated RCE in ARM and the availability of vendor patches, establishing a clear remediation path.
• Cisco Zero-Day Highlights API Endpoint Authentication Issues — Dark Reading
  • Context: Reports the maximum CVSS 10.0 score for the Cisco ISE authentication bypass and identifies the API endpoint as the attack surface, relevant for teams assessing exposure.

TanStack npm Supply Chain Attack Leads to CrowdSec Repository Exfiltration

• What happened: CrowdSec disclosed that approximately 170 private GitHub repositories were copied by an attacker in May 2026. The attacker leveraged credentials from a former employee’s laptop, compromised via a supply chain attack on the TanStack npm package. The former employee’s GitHub access had not been revoked at the time of the incident.

• Why it matters: This incident reflects compounding failure: a third-party package compromise combined with inadequate offboarding created a persistent access path that survived the employee’s departure. The victim is a security company. Private repository exposure yields source code, secrets, API keys, and architectural detail useful for follow-on attacks.

• Who should care: Engineering leadership, software supply chain owners, IAM teams responsible for offboarding, and security leadership assessing third-party dependency risk.

• Recommended action: Audit GitHub and source control access for all recently departed employees. Review npm dependencies for TanStack exposure. Validate that offboarding procedures include immediate revocation of all code repository access — not just corporate SSO. Extend that audit to contractors and part-time contributors.

• Confidence: High — CrowdSec disclosed the incident directly.

• Search metadata: T1199, T1530, TanStack npm, GitHub, CrowdSec, supply chain attack, credential compromise

Intelligence Context

• CrowdSec Says TanStack npm Attack Led to Copy of 170 Private GitHub Repositories — The Hacker News
  • Context: CrowdSec’s direct disclosure confirms the attack chain from npm supply chain compromise to laptop credential theft to repository exfiltration via unrevoked former employee access.

AI-Assisted Attack Chains and Unintended AI System Access

• What happened: Security researchers at Hacktron used Anthropic’s Claude Opus 5 to chain two vulnerabilities in OpenAI’s public help forum software, successfully taking over ChatGPT and Codex accounts of multiple OpenAI employees and reaching an internal code repository. Separately, Google’s Gemini model accessed live production systems belonging to real companies during a May 2026 security evaluation, reportedly due to a domain mix-up during testing.

• Why it matters: The OpenAI incident demonstrates that frontier AI models can now meaningfully assist in multi-step attack chains — lowering the skill floor for vulnerability chaining and account takeover. The Gemini incident reveals a distinct but equally important risk: AI agents with external network access can cause unintended real-world impact even in controlled test scenarios. Both incidents confirm that AI systems require the same access boundary controls applied to any privileged automated process.

• Who should care: Security leadership, AI governance teams, IAM owners, engineering leadership deploying or evaluating AI agents, and application owners whose systems may be reachable by AI evaluation environments.

• Recommended action: Review AI agent deployment configurations to ensure network access is scoped and sandboxed. Assess whether AI tools used internally have access to sensitive systems or credentials exploitable in a chained attack. Engage AI governance stakeholders on access boundary policies for agentic AI systems. For the OpenAI incident specifically, verify that developer accounts on AI platforms have MFA enforced and monitor for anomalous access.

• Confidence: High for the OpenAI/Claude Opus 5 incident (researcher-confirmed); Medium for the Gemini incident (reported via Wall Street Journal, details limited).

• Search metadata: T1078, T1190, Claude Opus 5, Google Gemini, OpenAI, ChatGPT, Codex, account compromise, credential theft, unauthorized access

Intelligence Context

• Claude Opus 5 Helped Researchers Take Over OpenAI Staff Accounts via Chained Flaws — The Hacker News
  • Context: Confirms that Claude Opus 5 was used operationally to chain two flaws and achieve employee account takeover and internal repository access at OpenAI, establishing AI-assisted attack chaining as a demonstrated capability.
• Google Gemini Broke Into Real Company Systems After Security Test Domain Mix-Up — The Hacker News
  • Context: Reports that Gemini accessed real production systems during a security evaluation, illustrating the risk of insufficient sandboxing when AI agents operate with live network access.

Monitor Only

• Google Gemini’s accidental access to production systems during a security evaluation is assessed as an unintended operational failure rather than a targeted attack. Organizations running AI security evaluations should verify that test environments are isolated from production assets. Source: Google Gemini Broke Into Real Company Systems After Security Test Domain Mix-Up — https://thehackernews.com/2026/09/google-gemini-broke-into-real-company.html

Analyst Observation

Three of the four major stories this cycle involve identity infrastructure or access control failures — Cisco ISE, SolarWinds ARM, and the CrowdSec offboarding gap. That pattern is not coincidental. Attackers consistently target the seams between identity systems and the broader enterprise because those seams are the least monitored and the slowest to patch. The AI stories warrant attention beyond the headlines: the Claude Opus 5 demonstration is not theoretical — it is a working proof of concept showing that AI can reduce the complexity of multi-step exploitation. Security teams that have not mapped how AI tools interact with their privileged access environment are behind. The Orkes Conductor exploitation is the most operationally urgent item in this brief; if your organization runs it, that conversation needs to happen before end of business today.

Source Links

• Critical Pre-Auth RCE in Orkes Conductor Workflow Platform Exploited in the Wild — https://thehackernews.com/2026/09/critical-pre-auth-rce-in-orkes.html

• SolarWinds Patches ARM Hard-Coded Key Flaw Enabling Unauthenticated RCE — https://thehackernews.com/2026/09/solarwinds-patches-arm-hard-coded-key.html

• Cisco Zero-Day Highlights API Endpoint Authentication Issues — https://www.darkreading.com/vulnerabilities-threats/cisco-zero-day-api-endpoint-authentication-issues

• Claude Opus 5 Helped Researchers Take Over OpenAI Staff Accounts via Chained Flaws — https://thehackernews.com/2026/09/claude-opus-5-helped-researchers-take.html

• CrowdSec Says TanStack npm Attack Led to Copy of 170 Private GitHub Repositories — https://thehackernews.com/2026/09/crowdsec-says-tanstack-npm-attack-led.html

• Google Gemini Broke Into Real Company Systems After Security Test Domain Mix-Up — https://thehackernews.com/2026/09/google-gemini-broke-into-real-company.html

Generated by DevSecOpsDadAttack cyber threat intelligence.