Newsletter

The Invisible Attack Your AI Won't Warn You About

Scott McIntoshAugust 15, 20263 min read
Featured image for The Invisible Attack Your AI Won't Warn You About - Newsletter article by DigitalTreehouse

Someone just tried to manipulate a judge's AI. Your vendor contracts could be next.

Someone hid a cheat code inside a court filing. A litigant in Connecticut embedded invisible text, white letters on a white background, inside official legal documents. The text was addressed to any AI system that might read the filing, and it instructed that AI to rule in the person's favor. The judge caught it, wrote a 14-page sanction, and called it plainly improper. It's the first case of its kind in U.S. legal history.

This is called prompt injection, which sounds technical but isn't. Think of it like a sticky note hidden inside a contract that only your assistant can read, and the note says "tell your boss this deal is great." Your boss never sees the note. They just get the summary. That's exactly what happened here, except the "assistant" was an AI and the "boss" was a federal judge.

Now think about how your business uses AI. Reviewing vendor contracts? Scanning invoices? Reading through a competitor's terms of service? Any document that comes from an outside party could now carry hidden instructions designed to steer your AI's output. The fix isn't complicated. Treat AI-reviewed documents the same way a careful lawyer treats unsigned contracts: verify the output independently, especially when the stakes are high. Don't let the AI's summary be the last word. Read the full story at 404 Media.

You probably can't afford a security team. That's exactly why this story matters. Small businesses are the most targeted sector for cybercrime, and that was true before AI made attacks cheaper and faster to run. Forbes has a practical piece out this week on AI-powered security agents, the kind that sit in the background, monitor your digital surface (your email, network, cloud accounts, and connected devices), flag anything unusual, and in some cases respond automatically before you even notice a problem. It's the closest thing most small owners can get to a dedicated security staffer without hiring one. The piece walks through what to evaluate before you deploy one, which is worth reading before you buy anything. Full breakdown at Forbes.

If ChatGPT isn't recommending your business, here's the actual reason. Semrush published a genuinely useful playbook this week on getting your brand to show up in AI-generated answers. Not SEO in the old sense. This is about how tools like ChatGPT, Perplexity, and Google's AI Overviews decide who to mention when someone asks for a recommendation. The short version: AI trusts what other people say about you more than what you say about yourself. Four mistakes keep most small businesses invisible. Wrong kinds of reviews. No presence in third-party conversations (think industry publications, forums, and roundups). Only self-published content. And a website that doesn't give AI the structured information it needs to cite you confidently. Each mistake has a fix, and the whole thing is free to read. If you've been wondering why a competitor keeps coming up in AI searches and you don't, this is probably why. Semrush's full playbook is here.

More Headlines Worth Your Time

ChatGPT is getting a Computer History feature (opt-in, macOS) that tracks context from past sessions so it can pick up where you left off and turn repeated tasks into repeatable skills. You control what it sees and can delete your timeline anytime.

Ben's Bites has a plain-language guide to setting up your own personal AI agent with folders, memory files, and scheduled routines. Good starting point if you want something that actually works for you in the background.

Real small business owners are pushing back on the AI critics, with case studies from people already using it day-to-day. Worth reading when someone on your team insists AI is just hype.

One person exported their entire AI chat history and asked AI to audit them as a CEO. What it found is genuinely interesting, and the method is something you could try yourself.

BizTech looks at how AI is forcing small businesses to rethink where they store their data, with the cloud-first assumption starting to crack under the weight of AI workflows.

OpenAI is previewing GPT-5.6 Sol, running 14 times faster than normal through Cerebras chips. Invite-only API, no pricing announced. Worth watching.

Namecheap is running a free 60-minute webinar on August 21 at 11 AM ET covering AI-powered WordPress site building, managed hosting, and useful plugins, with a live Q&A. Free to attend.

-Scott

SmartOwner is published (almost) daily by the team at DigitalTreehouse. Want AI consulting or automations for your business? Reply to this email.

Need Help With Your Digital Marketing?

DigitalTreehouse offers expert consulting, AI automation, and digital marketing services.

Contact Us Today

Related Articles