If you’ve been following along with our ongoing saga of “AI assistants behaving badly,” you might remember our earlier breakdown of MoltBot—later ClawdBot—now rebranded as OpenClaw, because apparently every AI needs a dramatic stage name evolution. (If you missed that post, go check out Everyone Loves MoltBot… Except Your Security Team. It’s basically the prequel where we warned you this sequel was coming.)

OpenClaw has become the darling of the internet: a personal AI agent that organizes your life, automates your tasks, and occasionally acts like a golden retriever hopped up on energy drinks. People adore it. Businesses deploy it. Teenagers use it to write essays. And somewhere out there, a cybercriminal is rubbing their hands together like a cartoon villain because OpenClaw is basically a buffet of security flaws wrapped in a friendly UI.

Small businesses don’t need a six‑figure security budget, a SOC full of hoodie‑wearing analysts, or a PhD in “Advanced Threat Matrix Quantum Zero‑Trust Blockchain.” You just need a handful of free (or cheap) things that actually move the needle — and won’t make your accountant cry.

AI is getting smarter, faster, and—unfortunately—way more interesting to hackers. And while most people are using ChatGPT to write emails, summarize meetings, or ask it why their houseplants keep dying, attackers are busy trying to trick AI systems into coughing up sensitive data through something called prompt injection.

To help stop that, ChatGPT now has a shiny new feature: Lockdown Mode. Think of it as the AI equivalent of putting your phone in airplane mode, locking it in a safe, and then burying that safe under a concrete slab. It’s not for everyone—but for the people who need it, it’s a game‑changer.

Smart TVs have quietly become one of the biggest data collectors in your home. If you own a TCL, Sony, Hisense, or any TV running Google TV, you’re not just streaming movies—you’re feeding a constant stream of behavioral data back to manufacturers, app developers, and advertisers.

If you’ve ever Googled a quick fix for a Mac issue, you already know the drill: search, skim, click the first result, and hope the instructions don’t break anything. Attackers know this too—and they’re exploiting it with a new twist. A recent campaign is abusing Claude LLM artifacts and Google Ads to trick macOS users into running Terminal commands that quietly install infostealers.