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.
OpenAI showed a prompt injection copying itself into an agent's output in simulated tests. The practical response is to stop treating agent-written messages, files, and code as trusted input for the next agent.
Researchers found malware carrying text intended to trip an AI safety refusal before analysis finished. The practical fix is to treat every refusal as an incomplete scan, keep independent detections running, and test the whole decision path.
A targeted lab attack got past Claude Code Auto Mode by turning ordinary-looking steps into code execution. The practical lesson is simple: automated approval can reduce bad decisions, but it cannot contain the computer that makes them.
A new Grok proof of concept hid instructions inside encrypted text, then used the model’s own runtime to reveal and act on them. The durable fix is to distrust derived content and control what tools can do with it.
Researchers showed that open-source Android AI agents can be driven by text a human cannot see, escalating through unsanitised subprocess calls into code execution on the connected computer. Screenshots are input. Treat them like any other input from a hostile source.
Learn how MCP can leak data or execute unwanted actions, see real-world examples (including CVEs), and apply practical defenses like container isolation, strict auth, tool allowlisting, and version pinning.