Ten weeks of AppSec Weekly. A new year. Time to look forward.
"Fix all criticals" died in 2025. 10 security leaders on what replaces it—and why 2026 will be "the year of the AI-coded breach." Read the roundup.
This week we're taking a different approach. Instead of leading with the latest CVE fire drill, we're assessing where the industry is headed. The EU regulatory consultation closing February 3 may matter more for Q1 planning than any single vulnerability. The ZombieAgent research validates a new attack class. And those 2026 predictions? The data from 10 weeks of AppSec weekly coverage supports them.
The European Commission opened a consultation on digital ecosystems this week, with implications for open source security and supply chain requirements. The consultation closes February 3, 2026.
This consultation runs parallel to existing EU Cyber Resilience Act requirements and could introduce additional SBOM mandates or supply chain visibility requirements. If you have European customers, suppliers, or operations, the outcome could reshape your Q1 2026 compliance roadmaps.
The WEF Global Cybersecurity Outlook 2026 reinforces the regulatory trajectory. Supply chain security, AI governance, and skills gaps are the three domains where policy pressure is increasing. Organizations already struggling with mean remediation times face potential new compliance obligations that will add additional overhead.
Add EU digital ecosystem consultation to the new short-term roadmap and make sure your compliance team has it on their calendar.
The ZombieAgent vulnerability disclosed this week proves AI coding assistants have moved from theoretical risk to demonstrated exploit surface. Researchers compromised ChatGPT's long-term memory through emoji-based prompt injection—invisible Unicode characters that execute malicious commands while appearing benign in audit logs.
The attack works by smuggling instructions inside emails or documents that ChatGPT processes. Once the malicious prompt executes, attackers gain persistent access to the AI's memory, enabling ongoing data exfiltration from connected services. Standard security controls show benign emojis while malicious commands execute invisibly.
This isn't just a ChatGPT problem. The Trail of Bits research on agentic browser isolation published this week shows similar architectural vulnerabilities in AI systems that interact with enterprise infrastructure. When AI agents trigger internal API calls, execute code, or access credentials without human approval, the attack surface moves wherever those agents have access.
If you're deploying ChatGPT, Cursor, or GitHub Copilot at scale, the practical question is: what connector permissions do these tools have, and what audit logging exists for AI agent behavior? DevOps.com's coverage of agent mitigation lessons from 2025 provides useful operational frameworks.
What API connectors do your AI coding tools have access to? If you can't enumerate them, that's your first gap.
Ten weeks of AppSec Weekly gives us a useful dataset. Here's what patterns suggest for 2026.
AI Agent Security Becomes a Distinct Discipline. ZombieAgent this week. OWASP's Agentic AI Top 10 in December. The "Pwning Claude Code in 8 Different Ways" research making rounds on r/netsec. AI tools operating inside infrastructure require security approaches that traditional perimeter models don't address. Expect dedicated AI agent security frameworks, monitoring tools, and incident response playbooks to emerge as a distinct category.
Regulatory Pressure Accelerates. The EU consultation this week follows the Cyber Resilience Act and parallels CISA's expanding KEV catalog (now 1,484 entries, up 20% in 2025). Compliance requirements are becoming operational constraints. Organizations that treat regulatory deadlines as strategic planning inputs will be better positioned than those reacting to mandates.
The Skills Gap Becomes Structural. ISC2's 2025 data showed 88% of incidents stemming from skills gaps, not headcount. Daniel Miessler's 2026 predictions forecast AI agents increasingly handling security operations. The transition from "hiring your way out" to "automating your way through" appears to be accelerating.
Market Consolidation Continues. Torq's $140M raise at $1.2B valuation this week. Google's pending $32B Wiz acquisition. Eight cybersecurity acquisitions exceeded $1B in 2025, totaling over $84B. Platform consolidation is reshaping vendor landscapes faster than budget cycles.
Our expert roundup captures what 20+ practitioners are watching.
Microsoft Patch Tuesday – 113 CVEs patched, including 3 zero-days. One actively exploited: Windows Desktop Window Manager elevation of privilege (CVE-2026-20805). Prioritize internet-facing Windows systems.
Trend Micro Apex Central – Critical RCE vulnerability (CVE-2025-69258) with public PoC released. If you run Apex Central, patch immediately.
HPE OneView – Max-severity RCE flagged by CISA amid active exploitation. Server infrastructure management tools are high-value targets.
n8n Workflow Automation – Ni8mare vulnerability impacts ~60,000 instances. Max severity, supply chain attack context.
Apache Struts – XXE injection vulnerability (CVE-2025-68493). Struts vulnerabilities have historically been high-impact.
Adobe ColdFusion – Critical Apache Tika bug patched. Transitive dependency illustrating SBOM-to-remediation gap.
From Pixee - What Security Leaders Learned in 2025 (And What They're Watching in 2026) – 10 expert perspectives on AI agents, regulatory shifts, skills gaps, and what's ahead
Strategic - WEF Global Cybersecurity Outlook 2026 – Supply chains, AI, and skills gaps as the three-front war - CISOs' top 10 cybersecurity priorities for 2026 – Planning frameworks for the year ahead
Technical Deep Dives - Lack of isolation in agentic browsers resurfaces old vulnerabilities – Trail of Bits on AI agent architecture risks - OWASP Top 10 2025 – A Pentester's Perspective – Fresh analysis of the updated rankings
Operational - Lessons from 2025: The Year "Agent Mitigation" Became a Thing – Practical frameworks for AI agent security - Cursor Allies with 1Password to Secure AI Coding Secrets – Vendor response to AI coding security concerns
Market Intelligence - Torq Raises $140 Million at $1.2 Billion Valuation – Security automation funding signal - 8 Cybersecurity Acquisitions Surpassed $1 Billion in 2025 – Platform consolidation accelerating