Exploited Vulnerabilities
CISA's Apr 20-24 KEV additions made exposure validation the top operational theme. The affected surfaces include Cisco Catalyst SD-WAN Manager, PaperCut NG/MF, JetBrains TeamCity, Microsoft Defender, Marimo, D-Link DIR-823X, Samsung MagicINFO and SimpleHelp. The practical task is to prove whether these products exist, whether they are reachable, and who owns remediation exceptions.
- Use CISA KEV dates and due dates as the review queue, but enrich them with internal asset ownership and business criticality.
- Cisco SD-WAN entries deserve special handling because CISA paired KEV additions with emergency directive and hunt guidance.
- Remote support and administration tools should be checked for internet exposure, weak segmentation and missing audit logs.
- Known ransomware-use flags on PaperCut and TeamCity should push backup, credential and lateral-movement review alongside patch status.
CISA KEV additions span Cisco SD-WAN, PaperCut, TeamCity and Microsoft Defender
CISA added exploited vulnerabilities across network management, developer infrastructure and endpoint security products during the week.
Cisco SD-WAN risk is treated as emergency exposure, not routine patching
CISA's Cisco SD-WAN emergency directive and supplemental hardening guidance make the SD-WAN entries a control-plane risk item for network owners.
AI & Exposure Management
Microsoft and Unit 42 both emphasized that AI changes the speed and scope of defensive work. Microsoft focused on AI-assisted vulnerability discovery, exposure management and rapid detections. Unit 42 showed why autonomous cloud attack testing and agentic workflows force defenders to reason about privileges, reachable data and guardrails before incidents happen.
- Treat AI risk as an exposure-management problem first: assets, code, permissions, internet-facing surfaces and hygiene.
- Separate AI-assisted discovery from automated exploitation claims unless a primary source supports the claim.
- For cloud and agentic systems, map what an automated workflow can read, change, execute and exfiltrate.
- Detection engineering should include model-assisted speed, but human review remains necessary for priority and context.
Microsoft: AI compresses vulnerability discovery and exploitation timelines
Microsoft connected AI model capabilities to vulnerability discovery, exploit chaining and the need for faster exposure reduction.
Unit 42 cloud AI research turns agent privilege into a design question
Unit 42's autonomous cloud offensive research is useful as a defensive prompt: know the privileges, reachable data and detection gaps for agentic workflows.
Identity, SaaS & Browser Control Planes
Identity risk moved beyond sign-in logs. Microsoft highlighted remote IT worker detection across Workday, DocuSign, Teams, Zoom, Webex and Microsoft 365. Unit 42's April bulletin added browser extensions and AI-agent input handling as identity-bearing surfaces where ordinary-looking user workflows can become data access paths.
- Coordinate HR, identity and security operations around external applicants, new hires and suspicious SaaS activity.
- Review browser extension controls with the same seriousness as third-party SaaS and endpoint software.
- Monitor external collaboration and helpdesk workflows for unusual access requests, remote-control tooling and data staging.
- For AI agents, treat web pages, documents and database content as potentially active instruction sources, not passive text.
Microsoft publishes detection strategies for infiltrating IT workers
The guidance focuses on finding suspicious behavior before and after onboarding by using SaaS and identity telemetry.
Unit 42 highlights browser extensions and prompt injection as AI-era trust issues
Unit 42's April bulletin described privileged browser extensions and indirect prompt injection as areas where data access and instruction handling blur together.
Software Supply Chain
Unit 42's npm research and FortiGuard's ActiveMQ signal both point to a practical supply-chain lesson: trust paths are exploitable when developer credentials, package lifecycle scripts, CI/CD secrets and middleware management interfaces are not constrained. The defensive priority is to reduce blast radius before one compromised component becomes an environment-wide incident.
- Disable or tightly govern package lifecycle scripts where build pipelines do not require them.
- Rotate exposed developer, npm, GitHub and cloud tokens when package compromise is plausible.
- Audit repositories for unauthorized workflow files, unexpected package version bumps and new public repositories.
- For ActiveMQ, restrict management interfaces and monitor abnormal Jolokia or Java child-process activity.
Unit 42: npm package attacks now emphasize credentials and CI/CD persistence
Unit 42's April 24 research describes npm compromise patterns that target tokens, package publishing rights and GitHub Actions workflows.
FortiGuard publishes ActiveMQ RCE threat signal for CVE-2026-34197
FortiGuard's April 20 signal and CISA KEV data align on ActiveMQ as a remediation priority, with management-interface restrictions and monitoring called out as useful controls.