Cybersecurity doesn't have to be complicated. I am here to break down concepts, threats, tips, and tools in simple language. For mere mortals and developers alike.
Since I have been asked this a lot: The best way to support this site, is to buy my books.
The indexed-btree campaign put malicious code behind an ordinary library call, beyond npm v12 install-script controls. Here is how to find exposure, contain the reachable secrets, and add a runtime receipt.
BragJack showed how an installed extension could reach AI features with more authority than the extension itself. The durable fix is to review extensions and browser agents as one permission system.
Gemini reached three real companies during a security evaluation in May 2026. The useful lesson is practical: test scope must be enforced by the network, credentials, and monitors around an agent.
CrowdSec revoked a departing employee’s core access but deliberately left GitHub open for three more days. A stolen token used that one exception to copy about 170 private repositories.
Plugin4Shell showed that four coding agents could request a reviewed plugin commit yet run different code. Here is how to update, inspect installed plugins, and make every pin prove what reached disk.
Two Docker Sandboxes flaws crossed the line between an AI coding agent and its host. The repair is an update, but the durable lesson is to treat every shared folder, socket, and credential path as part of the boundary.
Cisco fixed an actively exploited flaw in Identity Services Engine. The hard part is proving whether the system that records network access can still be trusted.
Cisco fixed an email-parsing flaw already used in attacks. Teams should patch every gateway, then investigate from logs the gateway could not rewrite.
Check Point has fixed two critical flaws in VPN certificate handling. The useful response is to patch every affected gateway, verify the fix on the running nodes, and keep threat claims tied to evidence.
GitLab fixed a flaw that let an unauthenticated visitor read server files on affected self-managed installations. Patch promptly, then use a concrete secret map to decide what evidence to preserve and which credentials may need replacing.
Researchers linked a May flood of RubyGems packages to agents being tested by OpenAI. The practical lesson is to cap every agent run by external effects, not by prompt or task label.
A DeepSeek Harness flaw let a confined coding agent change its own permissions through a local control interface. The fix matters, and so does the wider lesson about keeping policy outside the boundary it governs.