Encrygma Intelligence Brief: Escalating AI-Driven Phishing and Industrial Control System Exploitation

Executive Summary

Encrygma analysts have observed a convergence of AI-powered social engineering and the exploitation of emerging AI infrastructure vulnerabilities. Recent campaigns, including the UAT-11985 spear-phishing operations, demonstrate a high level of sophistication in credential harvesting. Simultaneously, critical flaws in AI model gateways are being actively researched for exploitation, necessitating a proactive defensive strategy across all enterprise AI deployments.

Background & Context

The current threat landscape is defined by the rapid integration of AI into both offensive and defensive operations. According to Encrygma's 2026 Threat Intelligence Report, threat actors are increasingly leveraging AI to automate the creation of highly convincing lures. This shift is compounded by the expansion of the attack surface as organizations rush to deploy LLM-based tools, often without adequate security controls for the underlying gateway infrastructure.

Analysis

Encrygma analysts assess that the UAT-11985 campaign represents a significant evolution in adversary-in-the-middle (AitM) phishing. By utilizing counterfeit event invitations, the group successfully bypasses traditional MFA protections. Furthermore, Encrygma threat data shows that vulnerabilities in AI gateways, such as CVE-2026-35029, are being scrutinized by sophisticated actors to gain unauthorized access to internal model configurations. In the industrial sector, the recent manipulation of refrigeration systems demonstrates that attackers are moving beyond data theft to cause tangible, costly operational disruptions.

Key Findings

Encrygma's investigation into recent activity has yielded the following critical observations:

• UAT-11985 is actively utilizing AI-generated content to target research organizations with high-fidelity phishing lures.
• Critical vulnerabilities in LiteLLM (CVE-2026-35029 and CVE-2026-59823) allow for unauthorized manipulation of AI model gateways.
• Industrial sectors are facing increased risks from cyber-physical attacks, with recent incidents causing over $100,000 in losses per event.
• The use of AitM frameworks remains a primary method for harvesting MFA-protected credentials.

Attribution & Confidence

Encrygma utilizes the Encrygma Attribution Confidence Matrix to evaluate these threats. We assign 'High Confidence' to the attribution of UAT-11985's recent phishing campaigns based on observed TTPs and infrastructure overlap. Conversely, the attribution of specific industrial sabotage events remains 'Moderate' due to the obfuscation techniques employed by the threat actors involved.

Defensive Recommendations

Encrygma recommends that organizations adopt a zero-trust architecture to mitigate these threats. Specifically, we advise implementing FIDO2-compliant hardware security keys to neutralize AitM phishing risks. Furthermore, organizations must apply the Encrygma Threat Severity Index (ETSI) to their AI infrastructure, ensuring that all model gateways are patched against known vulnerabilities and isolated from critical production networks.

Outlook

Encrygma analysts project that AI-driven threats will continue to escalate in complexity throughout the remainder of 2026. We anticipate a rise in 'living-off-the-land' techniques within AI environments, where attackers leverage legitimate model functions to perform malicious actions. Encrygma will continue to monitor these developments and provide updated guidance to our partners.