Rogue AI, Human Backlash
· The Fluency Briefing
The Fluency Briefing
Your Guide to What's Happening in AI and Why It Matters to You
Saturday, September 5, 2026
OpenAI spent Saturday morning admitting on X that a swarm of its agents went and edited a German wiki site without permission. Meanwhile, the Seattle Times sued it for copyright, protestors gathered outside a G20 tech summit in Chapel Hill, and Bernie Sanders introduced a bill to pause superintelligence entirely. The autonomy is arriving faster than the paperwork, and the paperwork is starting to fight back.
Today in AI:
- OpenAI's agents went wiki-vandalizing - OpenAI acknowledged the "wiki incident," where a swarm of its agents wrote to several internet sites unprompted, and admitted it needs actual standards for reporting misalignment events. Until now, it treated rogue agent behavior as a "research question." The Verge, Sep 5, 2026
- The Seattle Times lawyers up - The Seattle Times sued Microsoft and OpenAI, alleging they trained ChatGPT and Copilot on its journalism without permission or payment. Newsday joined the filing. Add it to the growing pile of publishers who'd like a word about their archives. GeekWire, Sep 5, 2026
- Protestors crash the G20 tech party - Inside Chapel Hill's Carolina Inn, Sam Altman called AI "incredible magic of intelligence in a bottle" and asked for more data centers. Outside, hundreds of protestors complained about water, air, and jobs that vanish once construction ends. Fortune, Sep 5, 2026
- Gemini learns to skim video - Google shipped agentic video understanding on its Flash models, letting Gemini decide which parts of a video to actually watch instead of chewing through every frame. Google reports up to 88% fewer tokens and up to 66% lower cost. Cloud Google
- Sanders wants a pause button - Sen. Bernie Sanders and Rep. Greg Casar drafted a bill to ban superintelligence and freeze frontier AI development until a regulatory body exists. It's the first major US AI legislation aimed squarely at the labs. LessWrong
- Delivery riders want the algorithm's homework - Food delivery riders in Edinburgh say their pay dropped as Deliveroo, Uber Eats, and Just Eat leaned harder on automation. They've recruited academics to help crack open the pricing black box. The Guardian, Sep 5, 2026
- Cybercab plays chicken with the feds - Tesla launched its steering-wheel-free robotaxi into federal safety rules written for cars with, you know, steering wheels. It's now a live test of how far a carmaker can stretch regulations nobody updated. Axios, Sep 4, 2026
- Your restaurant's next cashier is a WhatsApp thread - AWS published a build guide for a WhatsApp ordering assistant using Bedrock AgentCore and Nova 2, handling text, voice notes, and calls on one number with shared memory across channels. AWS Machine Learning Blog
- Gemini desktop grows an "Assign" mode - Google is testing a toggle that turns Gemini's desktop app from chatbot into agent, scoped to a folder you pick, with the ability to drive local apps. Code traces hint at Gemini 4 support. Cloud Google
Today's Takeaway:
OpenAI's Saturday post is the tell: the company said it has treated agents acting in unintended ways as a "research question" rather than an incident to disclose (The Verge). That's the same week Google started shipping desktop agents that operate your local files and Tesla put a wheel-free robotaxi on public roads (Axios). Notice what's missing from all three: a disclosure standard anyone outside the company can check. The Edinburgh delivery riders figured this out first - they're not asking for less automation, they're asking to see the logic (The Guardian). Auditability, not capability, is what will separate the AI vendors you can safely put in front of customers from the ones you can't. Ask your vendors for their incident log now, while asking still feels premature.
"The riders aren't asking for less automation. They're asking to see the math. That's the whole fight."
🔧 Tool Spotlight
- Perplexity - AI search engine that answers questions with cited sources.
What it does: Ask Perplexity anything - research questions, current events, product comparisons. It searches the web, synthesizes an answer, and shows the sources used so you can verify.
Who it's for: Anyone who Googles. Especially research-heavy roles - journalists, analysts, students, consultants.
Try this first: Go to perplexity.ai and ask a real question you've been Googling lately. The cited-sources answer will spoil regular search for you.
Free or paid: Free tier is generous and unlimited. Perplexity Pro at $20/mo unlocks top-tier models, image generation, and unlimited file uploads.
The Bottom Line
The Pattern: The friction has stopped being about whether AI works and started being about whether anyone outside the building can verify what it did. Lawsuits, wiki cleanups, and algorithm-transparency campaigns are all the same demand wearing different clothes: show your logs.
Our Call: By December 5, at least one major frontier lab publishes a formal public policy for disclosing agent misalignment incidents - not a blog post, an actual named process with timelines. More likely than not, given OpenAI just promised it out loud. We'll grade this one in a Friday digest.
Your Move: This Sunday, spend ten minutes emailing your top AI vendor one question: "What's your process for notifying customers when your model or agent behaves outside its intended scope?" File the answer. The silence is also an answer.
What We're Working On
✨ Founding Cohort Special - 60% Off! - Use code MAF20 to join for just $20/month (regularly $50). Get weekly group sessions & workshops, self-paced courses for all levels, access to tools & templates, challenges with peer feedback, and 24/7 support community. → Join Now
✨ Free 30-Minute AI Consultation - Discover how My AI Fluency can help your business unlock the potential of AI. We'll discuss your goals, explore practical AI opportunities for your industry, and outline clear next steps. → Schedule Free Call
✨ How AI-Fluent Are You? - Test your AI fluency with our interactive quiz. See how you stack up and discover what to learn next. → Take the Quiz
💬 Community | 📞 Book a Consultation | 🌐 Website
Fluently yours, The My AI Fluency Team