Threat Actors Actively Scan for MCP Servers and AI Assistant Credentials, Raising Access‑Control Risks
What Happened — The SANS Internet Storm Center reported a surge in automated scans that probe for exposed MCP (Managed Cloud Platform) servers and attempt to harvest credentials used by AI assistants. The scans target common ports and API endpoints, looking for mis‑configured authentication or default passwords.
Why It Matters for Compliance & Audit Readiness
- Continuous‑compliance programs require documented access‑control policies and evidence that credentials are protected against brute‑force or credential‑stuffing attacks.
- SOC 2 / CC 3.1 controls (Logical Access) must be demonstrably enforced and monitored; unchecked scanning can expose gaps that auditors will flag.
- Verisq’s SOC2_ACCESS_CONTROLS capability provides automated credential‑usage monitoring and policy enforcement evidence to satisfy audit requirements.
Who Is Affected — SaaS providers, AI‑assistant developers, and any organization running MCP‑type cloud workloads (primarily TECH_SAAS and CLOUD_INFRA sectors).
Recommended Actions
- Enforce MFA and strong password policies for all AI‑assistant service accounts.
- Deploy credential‑rotation automation and secret‑management tooling.
- Enable continuous log monitoring for failed login attempts and anomalous scan traffic; retain logs as audit evidence.
- Conduct a SOC 2 logical‑access control review and remediate any identified gaps.
Technical Notes — Attackers use widely available scanning tools (e.g., Nmap, Shodan) to enumerate open ports (22, 443, 8443) and then attempt credential stuffing against AI‑assistant OAuth endpoints. No specific CVE is cited; the risk stems from insecure configuration and weak credential hygiene.
Source: SANS Internet Storm Center