"Cloud services present a different security challenge than a typical on-premises application. While organizations cannot enforce execution-level controls within SaaS platforms, strong identification protocols along with access and data controls, can still reduce risk. -Danny Jenkins
From the ThreatLocker blog
Remote access, vendor relationships, and OAuth integrations all share a common theme:
What you need to know: Remote work brought with it an increased attack surface. Employees can access corporate resources from personal devices, home networks, and unmanaged environments using VPN and cloud applications that often implicitly trust authenticated users—an entry point cybercriminals are all too happy to exploit. MFA and SSO help verify identity, but they don't control what applications can execute or how far a compromised session can move. Applying Zero Trust to secure your remote workforce requires shifting from perimeter-based trust to continuous verification and least privilege.
Why it matters:Insider threats have increased with the rise in remote work, and remote workers are 3x more likely to accidentally expose company data than in-office employees. Attackers are also increasingly targeting user access because it is a legitimate entry point, and once they're in, overly broad permissions make lateral movement easier.
The bottom line: Identity authentication only verifies who a user is, not what they're doing. Assume a user can and will be compromised and build controls that prevent lateral movement, enforce least privilege, and monitor SaaS permissions to prevent remote access from becoming a weak link.
Zendesk compromise highlights increased attacks on SaaS relationships
What happened: In October 2025, Discord disclosed unauthorized access to customer support data hosted in Zendesk. There was no reported exploitation of a Zendesk software vulnerability, and Discord's core production infrastructure was not directly breached. Attackers accessed support tickets, internal communications, and attachments, including identity verification documents. Discord revoked access once the incident was confirmed and stated approximately 70,000 users were affected.
Why it matters: Modern organizations rely on SaaS platforms and vendors, and each integration introduces another identity with access to internal data and therefore another potential attack vector. Vendors are typically less secure but equally trusted, making them an ideal target for cybercriminals. Without the proper controls in place, a single vendor account compromise cascades into widespread exposure.
The bottom line: Your security posture is only as strong as your vendors' security, and attackers are more frequently targeting those trusted relationships rather than software vulnerabilities. Identity governance, vendor access visibility, least-privilege principles, and strong authentication enforcement are critical to any security posture.
Attackers gain persistent access without stealing credentials or triggering authentication
What's happening: ConsentFix represents a more advanced evolution of ClickFix-style social engineering where threat actors compromise a vendor's GitHub environment and obtain OAuth tokens associated with trusted integrations. These tokens can be used to access downstream customer environments at scale. By tricking users into authorizing malicious OAuth tokens to bypass authentication, attackers can interact with GitHub accounts and connected services without needing passwords or triggering MFA challenges.
Why it matters: OAuth is widely used to simplify authentication across SaaS platforms and developer tools, but excessive token permissions, long-lived tokens, and lack of visibility into app authorizations creates a significant blind spot. These token-based attacks bypass traditional controls because their activity appears authenticated. Victim organizations may experience exposure of CRM data, data exfiltration, unauthorized API access, intellectual property theft, and more.
The big picture: No single mitigation can eliminate this risk. Instead, it takes a layered approach to reduce exposure and limit attacker persistence. Monitor OAuth app approvals, restrict token scope, enforce least privilege, and audit third-party integrations.
Threats you need to know
From zero-days to AI endpoints, the pattern is clear:
Privileged systems without oversight become long-term footholds.
Maximum severity vulnerability in Dell RecoverPoint exploited as zero-day
What you need to know: Security researchers from Google disclosed the critical zero-day vulnerability (CVE-2026-22769) in Dell RecoverPoint for Virtual Machines that had been actively exploited since at least 2024. The flaw stems from hard-coded credentials tied to an “admin” user in the Apache Tomcat Manager instance. An unauthenticated attacker could gain OS-level access, upload a web shell, and execute commands as root. From there, attackers deployed the BRICKSTORM backdoor and later transitioned to GRIMBOLT.
Why it matters: This vulnerability enables direct root-level persistence without authentication, making it particularly dangerous when not closely monitored. Google Mandiant warns that the actor likely had significant time to establish long-term espionage access. UNC6201, the suspected China-nexus threat cluster, also used temporary virtual network interfaces to pivot into internal or SaaS environments, which they then deleted to evade detection.
The bottom line: Upgrade to version 6.0.3.1 HF1 or follow Dell’s migration guidance for affected versions. Dell also recommends deploying RecoverPoint for VMs within access-controlled internal networks only.
Exposed endpoints expand risk across LLM infrastructure
AI integration is a new attack surface
What you need to know: As more organizations deploy LLMs, they also rapidly expand the number of internal APIs, service connections, and automation endpoints that support them. Many LLM endpoints are built quickly for experimentation and left running with excessive permissions, static credentials, or minimal oversight, meaning exposure happens gradually, not due to a single failure. When these endpoints accumulate broad access and long-lived credentials, they become high-value attack vectors.
Why it matters: LLMs are designed to connect multiple systems within a broader technical infrastructure. With one compromised LLM endpoint, attackers can expose API keys, service accounts, and privileged automation workflows. A compromised endpoint can also enable prompt-driven data exfiltration, abuse of internal tool-calling permissions, and unauthorized access to cloud services or internal systems, all while appearing legitimate.
The bottom line:As AI adoption accelerates, endpoint privilege management will determine whether LLM deployments remain resilient or become high-speed pathways for lateral movement. Enforce just-in-time access, least privilege, and regular credential rotation to reduce the blast radius of exposed endpoints.
Last call to be part of Zero Trust World in Orlando
Join us for three days filled with:
Hands-on hacking labs, practical strategy sessions, and advanced technique training
Keynote sessions with former White House CIO Theresa Payton, Marcus Hutchins, and Adam Savage
Hack-it-to-win-it contest with a $5K custom-built PC up for grabs
