1-minute read
What a defender needs to know
Google Threat Intelligence Group examined an exposed staging server hosting exploit_agent, a framework that compares web-app code around patch releases, drafts exploits and executes them without human approval. GTIG found exfiltrated victim data and scripts for reconnaissance, brute-forcing and webshell installation. It identified French media and entertainment websites and education sites as particular interests from victim telemetry. The scripts represent available capabilities, while the report describes the framework’s automatic exploitation workflow.
For an exposed web app, a newly published fix helps attackers inspect the older version still running on your server. Inventory the internet-facing installations and their actual deployed builds, prioritize security updates when the patch ships, and retain reverse-proxy and application access logs across the release window. Inspect unusual probing of changed routes and subsequent outbound transfers or new web-served files. GTIG describes forensic traffic from unrestricted autonomous execution, but supplies no validated exploit_agent signature; investigate suspicious sequences in the context of each app.
The report separately cites exploited PaperCut flaws CVE-2026-81578 and CVE-2026-82078. PaperCut MF and NG first fixed builds for the affected 24.x, 25.x and 26.x lines are 24.1.10, 25.0.13 and 26.0.5. If you operate an exposed PaperCut management interface, apply the fixed build for your line and restrict its reachability to required management paths. Read the running version again to verify the update.