Posts

Two disclosures, one pattern. The AI tools adopted for productivity are now both an exfiltration surface and an instruction-injection surface.

A practitioner primer on protocol abuse: how data quietly leaves networks through DNS, ICMP, HTTPS/TLS, and the headers of TCP itself.

The first publicly available Mythos-class model. A look at capability gating, the two-tier release model, and what it signals for defenders.

A practitioner breakdown of the npm worm that sweeps environment variables on install and republishes itself through Trusted Publishing.

The largest DBIR ever published, read for signal. Where generative AI actually shows up in the data and where the fundamentals still rule.

Google caught the first AI-assisted zero-day in the wild. How attackers talked a model into helping, and why the method matters more than the exploit.

What Careful Adoption of Agentic AI Services means for non-human identity, privilege design, and compliance programs that haven’t caught up.

The BlackCat sentencings expose the trust every organization quietly extends to its incident response vendor, and how to structure it better.

CISA and NCSC published a joint analysis of FIRESTARTER, a backdoor that survives reboots and evades signature detection on Cisco edge devices.

Cisco compromised Claude Code’s persistent memory to deliver insecure guidance across projects, sessions, and reboots. What engineering teams should do.