A fact-based update for security and risk professionals, focused on how AI is reshaping the threat landscape and the defensive stack.

🔐 Core Security Intelligence

1) Cisco warns of active attacks exploiting zero-day in AsyncOS appliances

What’s new
Cisco has publicly confirmed that a maximum-severity zero-day vulnerability in the AsyncOS software used by Secure Email Gateway and Secure Email/Web Manager appliances is being actively exploited by sophisticated threat actors. The flaw allows attackers to achieve root access without authentication on affected devices, and there is currently no patch available. Cisco Talos attributes the campaign to an advanced persistent threat group linked to China.

Source:
Cisco Warns of Active Attacks Exploiting Unpatched Zero-Day in AsyncOS

Why it matters
AsyncOS appliances often serve as critical mail and web security gateways in enterprise environments. When these devices are compromised, attackers can intercept, manipulate, or degrade traffic that includes API calls, webhook alerts, or telemetry from AI systems, giving them an indirect vector to influence or surveil model deployment workflows and incident alerts.

Defenses

• Immediately isolate affected appliances from untrusted networks and limit access to management interfaces.
• Apply compensating controls such as strict ACLs, conditional VPN access, and additional MFA for admin consoles.
• Treat logs from these devices as high-value telemetry and feed them into SOC detection pipelines to catch lateral movement early.

Expert insight
Even though AsyncOS itself is not an AI product, its position at the perimeter makes its compromise an AI-relevant risk because it can serve as a conduit for adversaries to observe, intercept, or inject content into contexts where AI security tooling operates. Hardening core perimeter gear remains essential for safe AI operations.

2) Study shows WAFs ineffective against React2Shell exploits in modern environments

What’s new
A new benchmark study from Miggo Security reveals that many traditional Web Application Firewalls (WAFs) fail to stop exploitation of React2Shell (CVE-2025-55182) and similar unauthenticated remote-code execution vectors. The research highlights significant blind spots in how generic WAF rules are applied, particularly against threats that use synthetic and AI-crafted payloads designed to evade pattern-based detections.

Source:
Benchmark Study Finds WAFs Ineffective Against Latest React2Shell Exploit

Why it matters
Modern AI services and dashboards often expose web-based endpoints built on frameworks like React and Next.js. If perimeter defenses like WAFs are ineffective against sophisticated, obfuscated exploits, especially those that could be generated or refined with AI assistance, enterprise AI applications become vulnerable to compromise at the edge.

Defenses

• Adopt behavior-based web filtering and anomaly detection in addition to signature rules.
• Deploy runtime application self-protection (RASP) inside critical AI web services.
• Harden input validation and error handling in endpoints that expose model serving or AI orchestration functions.

Expert insight
This research reinforces that perimeter defense must evolve. Traditional WAFs tuned for static signatures will not suffice against payloads that are dynamically created or mutated by AI tools. Integrating upstream detection with downstream context, such as combining WAF logs with model activity, offers a better defense posture.

3) SonicWall patches exploited zero-day in SMA1000 after active exploitation

What’s new
SonicWall released patches for CVE-2025-40602, a local privilege escalation vulnerability in its Secure Mobile Access (SMA) 1000 appliances that has been exploited in the wild, often combined with other flaws to achieve unauthorized root access. Attackers targeted the Appliance Management Console, and SonicWall urges immediate updates to mitigate ongoing attacks.

Source:
SonicWall Zero-Day Privilege Escalation (CVE-2025-40602) Patched After Exploitation

Why it matters
SMA appliances provide secure remote access, including to cloud management portals, AI orchestration consoles, and internal admin interfaces. A compromise here directly elevates the risk that threat actors can pivot into environments hosting AI pipelines or datasets.

Defenses

• Apply the vendor’s security update immediately, and verify version compliance across all SMA 1000 units.
• Restrict management access via Zero Trust network controls and microsegmentation.
• Audit remote access logs for unusual connections or privilege escalations consistent with prior attack patterns.

Expert insight
Remote access infrastructure is a perennial target. When these appliances are weak, threat actors can bridge from perimeter compromise into high-value environments, including those supporting sensitive AI systems, without needing to breach deeper defenses.

⚠️ Updates & Follow-ups

React2Shell exploitation cases rising toward all-time high

Recent reporting confirms public exploits for React2Shell (CVE-2025-55182) have reached historically high levels of availability, meaning defenders should expect continued scanning and exploitation pressure against web apps using React/Next.js stacks.
Source:
React2Shell Fallout Spreads to Sensitive Targets as Public Exploits Surge

Actionable follow-up:

• Monitor threat intel feeds for new exploit signatures.
• Expand WAF and API protection to incorporate behavior baselines.

📊 At-a-Glance Summary

Categories: Cybersecurity News

Tags: AI Security, Enterprise Risk, Network Security, Perimeter Defense, React2Shell, Vulnerabilities