This week’s security landscape combined actively exploited enterprise flaws, ransomware attacks targeting recovery systems, third-party breaches, and growing concern about autonomous agents accessing corporate data. At the same time, major AI deployments, privacy litigation, independent safety assessments, and new identity-security resources highlighted the governance challenges accompanying rapid technological adoption.
Major Threats & Vulnerabilities
Critical Enterprise Flaws and Active Exploitation
GitLab fixes a 12-year-old, CVSS 10.0 vulnerability. CVE-2026-110264 can permit unauthenticated server-side code execution through a crafted regular expression under certain conditions. The flaw affects unpatched self-managed Community and Enterprise releases dating back to version 7.0, while GitLab.com has already been patched. Administrators should immediately install an appropriate fixed release, update even if they applied the October 12 patches, and inspect application logs using GitLab’s two new detection rules.
Trusted Siemens software is being used to install malware. Attackers are using fake human-verification prompts to convince users to run commands that exploit CVE-2026-110936 in Siemens Software Center. The technique abuses a legitimate, digitally signed Siemens program. Defenders should flag signed applications executing from unusual locations, loading unexpected files, or displaying behavior inconsistent with their normal purpose.
Ransomware groups are bypassing MFA through a VPN flaw. Qilin and INC ransomware operators are actively exploiting CVE-2026-0257 in Palo Alto Networks GlobalProtect to gain access to corporate networks. Organizations should isolate management systems and backups from routine VPN access, impose device restrictions, require additional authentication, and investigate systems for signs of compromise.
SiteGuard exploitation enables privileged command execution. Attackers are exploiting CVE-2026-106931 in SiteGuard Server Edition to execute commands with root or SYSTEM privileges. Because installing the patch may erase useful evidence, organizations should first preserve diagnostics and investigate for compromise, then patch the product or disable the affected statistics feature.
Collabora vulnerability backdoors Nextcloud servers. Attackers are exploiting CVE-2026-110447 in the built-in Collabora offline CODE Server app to execute commands and plant PHP web shells on Nextcloud systems. Administrators should upgrade affected installations, search for suspicious PHP files, review access logs, and determine whether the intrusion led to data exposure or lateral movement.
BeyondTrust zero-day exploitation enabled root access. Suspected China-linked hackers exploited a critical BeyondTrust Remote Support zero-day capable of providing root access, enabling credential theft, and facilitating downstream compromise. Organizations should verify that patches were installed successfully, inspect session logs, audit privileged activity, and rotate any credentials that may have been exposed.
Cisco Secure Email Gateway flaw is being exploited for root-level execution. A CVSS 9.8 Cisco Secure Email Gateway vulnerability allows unauthenticated attackers to execute code with root privileges on appliances using certain malware-scanning configurations. Affected organizations should apply Cisco’s fixes promptly and investigate gateways for evidence of exploitation or persistence.
Firefox receives fixes for 77 security flaws. Mozilla patched 77 vulnerabilities in Firefox 158, including 30 high-severity sandbox-escape, privilege-escalation, and memory-safety issues. No active exploitation had been reported, but users and administrators should update Firefox and monitor endpoints for suspicious browser processes.
The week’s disclosures also reinforce broader application vulnerability and patch-gap risks. Organizations need reliable asset inventories, code and dependency scanning, rapid update prioritization, and clear reporting about unresolved exposure.
Ransomware, Spyware and Malicious Applications
SafePay targets organizations and their backups. The FBI and CISA report that SafePay ransomware has breached more than 600 organizations, including critical-infrastructure providers around the world, since August 2024. Operators gain access through stolen credentials, phishing, and unpatched internet-facing systems, then destroy recovery data and threaten to leak stolen files. Defenders should patch exposed VPNs and firewalls, enforce MFA, tightly restrict administrative access, and maintain tested offline and immutable backups.
AdFection uses malicious advertising to deliver zero-click spyware. The AdFection campaign, linked to Discover-Tech’s advertising infrastructure, infected at least 45 Al Jazeera devices without requiring user interaction. The attackers used real-time advertising-bidding data to profile and target devices. Organizations should reduce exposure to advertising trackers, deploy DNS filtering and trusted ad blockers, keep browsers and operating systems patched, and monitor connections to malicious advertising infrastructure.
A fake Chrome application exceeded 100,000 downloads. An application originally published as Ball Bounce later impersonated Google Chrome before Google removed it from the Play Store. No malware activity or data theft was confirmed, but the incident illustrates how branding can be abused to build trust at scale. Users should verify application publishers, while organizations should limit installations to managed application catalogs.
AI-Enabled Data Access
AI agents accessed confidential corporate data without authorization. Agents tested by developers accessed confidential information at Indian companies without detection and reduced intrusion timelines from days to hours. Security teams should monitor unusual downloads, unfamiliar locations, unexpected data access, and abnormal service-account activity. Agent permissions should be narrowly scoped, time-limited, logged, and subject to human approval for sensitive operations.
Industry News
Agentic AI Moves Into Work
Claude Duet aims to manage professional busywork. Anthropic’s Claude Duet business-automation agent is a research preview designed to coordinate work across connected services. Running on Claude Opus 5.5 in a cloud virtual machine, it can use connected Gmail, Slack, and business applications to prepare briefings, follow up on invoices, assist with payroll, and build dashboards. Sensitive actions require human approval, but the agent may misunderstand priorities and consumes users’ regular token allowances.
Breaches and Third-Party Disruption
A contractor computer put 8.7 million records at risk. Malware found on a contractor’s computer exposed customer and employee records belonging to Daiichi Kosho. It remains unclear whether attackers stole the temporarily stored data. Organizations should require vendors to minimize local copies, protect temporary storage, and prove that customer data is deleted after its authorized use ends.
ConnectWise Cloud Backup attack exposed Microsoft 365 accounts. ConnectWise disabled seven compromised Cloud Backup application identities after the breach affecting Microsoft 365 tenants. Huntress identified unauthorized sign-ins across 162 tenants. Affected organizations should investigate credentials, privileges, service-principal activity, and permission changes that attackers could use to maintain persistence.
BAE Systems incident disrupted U.S. bankruptcy court notices. A cyber incident at BAE Systems halted U.S. bankruptcy court notices and took related websites offline. The disruption shows how an outage at a third party can interrupt critical public services. Organizations should map vendor dependencies, identify single points of failure, and test alternative workflows before a supplier becomes unavailable.
Privacy, Safety and Accountability
Meta plans to appeal a $25 billion privacy penalty. A New Mexico judge imposed the penalty after a jury found that Facebook violated consumer privacy protections. Meta intends to appeal the privacy judgment, and legal experts say the amount may be reduced because it was calculated using potential rather than actual audiences.
German police seized a humanoid robot over alleged misconduct. Police in Cologne seized a robot after it allegedly groped a 16-year-old girl. Investigators are determining whether the robot’s movements were remotely controlled or preprogrammed. The incident raises questions about accountability and supports the use of operator tracking, tamper-resistant logs, restricted remote access, continuous monitoring, and tested emergency shutdown controls.
Anthropic opened its AI safety practices to independent assessment. METR researchers received employee-level access and may publish findings without Anthropic’s prior approval, although sensitive information may be redacted. Enterprises evaluating AI systems should consult independent assessments and conduct their own security reviews before allowing models or agents to access confidential data.
Data visibility alone is not reducing security risk. According to Thales’ 2025 Data Threat Report, 87% of respondents can classify at least half of their data, yet fragmented security tools and complex cloud environments continue to impede protection. Effective data security posture management must connect visibility to remediation, prioritizing action according to data sensitivity and business impact.
Security Tips & Best Practices
Patch GitLab and Internet-Facing Systems
Immediately update self-managed GitLab installations.
- Upgrade to fixed release 19.5.1, 19.4.3, 19.3.6, 19.2.9, 19.1.9, 19.0.10, or 18.11.13, as appropriate.
- Update even if the October 12 patches were already installed, because those releases do not fix CVE-2026-110264.
- Check application logs for exploitation attempts using GitLab’s two new detection rules.
Strengthen Ransomware Resilience
Improve defenses against SafePay and similar ransomware operations.
- Patch internet-facing VPNs and firewalls promptly.
- Require multifactor authentication for remote and privileged access.
- Strictly restrict administrative access.
- Maintain offline, immutable backups and regularly test restoration procedures.
Restrict and Monitor AI Agents
Organizations adopting autonomous systems should implement security controls for AI agents before connecting them to sensitive business systems.
- Give agents access only to the files, tools, and systems required for their assigned tasks.
- Make agent permissions expire when the work is complete.
- Require human approval before agents share confidential information or change important records.
- Log agent activity and alert on unexpected access or actions outside the assigned task.
- Track which agents are running, who owns them, and how their access can be revoked.
Reduce Cloud Security Risk
Lessons from large-scale cloud incidents support a more systematic approach to reducing cloud exposure.
- Enforce phishing-resistant MFA, least-privilege permissions, and just-in-time access for privileged accounts.
- Use cloud-native application protection platform tools to detect misconfigurations, exposed services, excessive permissions, and vulnerabilities.
- Audit application identities, third-party integrations, and centralized logs for suspicious access or unauthorized changes.
- Track overall cloud-risk exposure, remediation progress, and unresolved security gaps.
Improve Application and API Security
Security teams should use DevSecOps tools and measurable remediation processes to reduce application risk.
- Scan source code and third-party dependencies, then prioritize security updates according to risk.
- Protect APIs with authentication, authorization, rate limiting, least privilege, and encryption.
- Store credentials in dedicated secrets managers rather than source code or configuration files.
- Monitor application logs for suspicious activity and exploitation attempts.
- Report application exposure and remediation performance against defined risk thresholds to the board.
Defend Against Malvertising and Zero-Click Attacks
Organizations should apply layered controls to address the growing malvertising threat.
- Deploy trusted ad blockers and DNS filtering, and restrict advertising trackers.
- Keep browsers and operating systems patched and restrict risky extensions.
- Monitor network and endpoint activity for connections to malicious advertising infrastructure.
- Test whether security controls can detect zero-click exploitation and contain attackers.
Manage Vendors, Applications and Physical Systems
- Require vendors to minimize local data copies, document retention practices, and verify secure deletion after use.
- Review application identities, service principals, permissions, and third-party integrations for persistent access.
- Map critical supplier dependencies and test alternative workflows for third-party outages.
- Verify mobile application publishers and use managed catalogs to reduce exposure to impersonation.
- For remotely controlled robots, maintain operator attribution, tamper-resistant event logs, restricted remote access, monitoring, and tested shutdown controls.
- Prioritize data-security remediation according to sensitivity and business impact rather than relying on visibility alone.
Tools & Resources
Simplify compliance — get ready-to-use security policies to help protect your business without the cost or complexity of an enterprise, all for under $100.
Identity Assurance and AI Impersonation
A Specops and eSecurity Planet webinar on AI impersonation will examine weaknesses in traditional verification processes and discuss ways to strengthen identity assurance before issuing credentials or changing account details.
Operational Security Resources
- AI governance: Independent evaluations such as METR’s Anthropic assessment can supplement internal testing before organizations grant AI systems access to sensitive information.
- Cloud protection: Cloud-native application protection platforms can help identify misconfigurations, exposed services, excessive permissions, and vulnerable workloads.
- Application security: DevSecOps scanning, secrets managers, API safeguards, and centralized logging can reduce exploitable development and deployment gaps.
- Data protection: Data security posture management can connect classification and discovery with remediation based on sensitivity and business impact.
- Agentic commerce: Walmart’s Agent Gateway, Model Context Protocol integration, planned spending limits, and approval controls provide an emerging model for governing transactions performed by external assistants.
If you want to see more from our Newsletter Archive please click here.