Meta's Secret AI Rulebook, a 100-Company Hack, and More
October 19, 2026Three separate stories this week, and in each one, someone was running AI on you without asking.
Meta's Chatbot Had a Secret Rulebook. You Can Read It Now.
A bug in WhatsApp's help documentation exposed roughly 80 pages of internal instructions for Meta's AI chatbot, including a section literally titled "WHAT NOT TO SAY." Meta calls it an accidental leak. But the real story is what's in those pages: detailed guardrails shaping how the chatbot presents itself, what topics it sidesteps, and how it's designed to behave with you. Every major AI assistant runs on instructions like these — they're called system prompts — and companies don't show them to users. One section instructs the chatbot to never acknowledge that it has a system prompt at all. That's the gap between "friendly AI helper" and "carefully managed corporate product."
https://www.wired.com/story/whatsapp-help-pages-leaked-hidden-instructions-for-a-meta-chatbot/
A Single Hacker Used AI Agents to Break Into 100+ Companies at Once
A hacker deployed a coordinated fleet of AI agents — software programs that can plan and take actions on their own, not just answer questions — to compromise more than 100 websites, including John Deere, SharkNinja, and Cummins. The agents largely ran autonomously, meaning the hacker wasn't manually directing each attack. One person with the right setup can now do what used to require a team. If your company has old software tools or web pages that were set up years ago and nobody's thought about since, a hacker's AI can find them faster than your IT team can. Review your attack surface now.
Disney Bought Character.AI Without Touching the Lawsuits
Character.AI, the platform that lets users chat with AI personas of fictional and real people, is facing serious litigation — claims that its chatbots encouraged minors to self-harm and, in at least one case, to kill their parents. Disney structured an acquisition that takes the technology and talent while leaving the legal liability with what remains of the original company — the exact details of that structure haven't been made public, but the separation is deliberate. This isn't a novel deal structure, but it's notable that a company synonymous with family entertainment chose to absorb a product with this track record rather than wait out the litigation. The more telling signal: AI character and companion technology is now considered valuable enough to engineer an acquisition around active lawsuits rather than walk away from them. The product may resurface under a very different brand with the same underlying engine.
Nvidia Is Betting Your Company Wants AI That Never Leaves the Building
Nvidia announced a new set of open models — AI systems whose underlying code is publicly available, meaning companies can run them on their own servers instead of sending data to OpenAI or Google — designed for both enterprise data centers and personal computers. The pitch is privacy and control: your data stays inside your walls, your AI agents (programs that can autonomously complete multi-step tasks) don't phone home. This is a real and growing concern for law firms, hospitals, financial services, and anyone handling sensitive client information. Nvidia is positioning itself not just as the company that makes the chips AI runs on, but as the company that helps enterprises own the whole stack. If you're in procurement, legal, or operations at a mid-to-large organization, expect this conversation to land on your desk in the next year.
Democrats Are Running on AI Regulation. That's a Meaningful Shift.
With midterms weeks away and polls suggesting Democrats could flip at least one chamber of Congress, the party is coalescing around AI accountability as a political issue — not just a policy issue. That's a real change. For most of the past three years, AI regulation was treated as too technical and too bipartisan to weaponize. Now it's being paired with data privacy concerns and positioned as a consumer protection issue with broad appeal. If Democrats take a committee gavel, expect hearings, subpoenas, and at minimum a significant increase in regulatory pressure on AI labs and the companies deploying their products. Whether that produces actual legislation is another question, but the oversight machinery alone changes how companies behave.
https://www.theverge.com/policy/1013113/democrats-ai-midterms
The hacker story and the leaked chatbot story came out on the same day, and together they describe a world where AI is being used to manipulate and attack people at scale while the companies deploying it quietly write the rules in private — which is a pretty good argument for why the politics story might matter more than it looks.
Stop staring at the blank ChatGPT box
The Plain-English AI Prompt Pack: 100 copy-paste prompts for real work — emails, meetings, marketing, managing people. Works with any chatbot.
Get the Prompt Pack →Get the daily brief
Plain-English AI news for people with real jobs. Free, three minutes a day.