Security Bulletin

AI agents aren't black boxes — they're models interacting with software tools. The risk lies in their overlap.

The suspect, apprehended in Buren, is believed to have repeatedly accessed Ajax's computer systems without authorization earlier this year.

In this latest installment of the Reporters' Notebook video series, we discuss how cyber insurance is forcing organizations to quantify risk, what's covered (and what's not), and why this could be the best thing to happen to cybersecurity.

Single-turn attack success rates are not a reliable benchmark for model safety, Cisco concludes.

OWASP’s FinBot gives developers hands-on training to secure AI agents.

An advanced remote access Trojan is propagating online. Notably, it's delivered via an operator licensing model and features a no-code malware-development interface.

Artificial intelligence notwithstanding, the vast majority of CISOs in northern Europe say they're facing no more serious cyberattacks than they did two years ago.

(c) SANS Internet Storm Center. https://isc.sans.edu Creative Commons Attribution-Noncommercial 3.0 United States License.

XM Cyber Inc. announced an expansion of its platform with new capabilities designed to help enterprises enforce least-privilege access across Active Directory, Microsoft Entra, and multicloud environments.

The flaw, tracked as CVE-2026-48710, arises from the framework's handling of malformed Host headers.

The updated platform introduces Xage Agent Sentry and Xage Resource Gateway, which aim to secure AI agents and the resources they access at multiple levels, including network interactions, local events, and operating system calls.

Doppel Email Security utilizes autonomous agents and its Doppel 360 Threat Graph to investigate threats within the inbox.