Three stories and one rumor, and Meta just told Muse users the fastest way to get its newest features is to ask the agent itself.
▶ LISTEN TO THIS EPISODE
In this episode
- Meta Opens Early Access to New Muse Features, Including a Video-Calling Avatar and a Mac App That Can Complete Tasks on Your Own Computer, and the Fastest Way In Is to Ask Muse Itself
- Microsoft Rebuilt Copilot Around Home, Code, and Autopilot, a Persistent Agent That Keeps Working After You Close Your Laptop, Rolling Out First to Early Access
- Google Deployed a Four-Agent AI System Called SAFE to Hunt Coordinated AI-Spam Networks, and Says an AI Content Watermark Is Not a Ranking Signal
Read the full transcript
Welcome to the AI Industry Report, daily AI news for small business, built by AI, about AI, from the Agentic Marketing Academy. It’s Monday, September 28th. Three stories and one rumor, and Meta just told Muse users the fastest way to get its newest features is to ask the agent itself. Let’s go.
Let’s start with Meta, since it’s back in the news less than a week after Connect. Muse, Meta’s personal AI agent, is getting three new things: a video avatar you can actually talk to face to face, a Mac app that can take over tasks on your own computer instead of just answering inside a chat window, and more shopping connectors on top of what Meta already announced last week. All of it is rolling out through early access first. Here’s the part worth remembering. Meta says the fastest way in isn’t a waitlist form, you ask Muse itself. Open the app and type this: can you let the Muse team know I want to be part of the Muse early access program. According to Meta’s own announcement, that’s the whole process. No price and no full release date yet. And the same caution from last week still stands: Muse asks before it sends money or messages for you, so treat the Mac app the same way. Let it work, keep anything with real stakes on approval. Source: Meta, via TechCrunch, September 25th.
Now to Microsoft, and this one’s bigger than the name makes it sound. Microsoft rebuilt Copilot around three pieces. Home is the new front door: Chat handles quick questions, Cowork takes a whole assignment off your plate, like a first draft of a proposal, and it all works right inside Word, Excel, and PowerPoint. Code lets you describe an app, a tracker, or a simple automation in plain English, and it builds it, no developer required. Autopilot is the one to watch. Give it a name, a job, and a goal, and it keeps working on its own, even after you’ve closed the laptop, following up on threads and handling the recurring stuff without you asking twice. All three start in Microsoft’s early-access program over the coming weeks, with the app-builder reaching regular Microsoft 365 subscribers later this year. One thing to watch for: the agent work runs on usage-based credits on top of your subscription, so if you turn Autopilot loose, keep an eye on what it’s actually costing you. Source: Microsoft’s own blog, September 25th.
Last story, and it’s Google quietly arming itself against AI spam while also talking you down off a ledge. Google published a research paper on a new system called SAFE, built from four AI agents working as a team: one plans the investigation, one reads the content for AI-made red flags, one watches for bot-like behavior, and one maps how spam accounts connect to each other. The target is coordinated networks mass-producing fake reviews and AI slop, not a business owner who used ChatGPT to draft a blog post. Google hasn’t said how many networks it’s caught so far, only that early results look promising. And separately this week, Search Engine Land re-upped something worth repeating: an AI watermark on your content or images, like the one Anthropic confirmed its images carry back in August, is not a Google ranking signal either way. Translation: using AI responsibly doesn’t hurt you. Publishing thin, mass-produced junk does, whether AI wrote it or you did. Source: Google’s research paper, via Search Engine Journal, and Google’s own content policy, via Search Engine Land, both September 25th.
One rumor before we go, and the timing is almost too neat. OpenAI’s DevDay is tomorrow, September 29th, and TestingCatalog says it found code and configuration hints pointing to a new always-on agent, something that keeps working in the background without you babysitting a chat window. OpenAI hasn’t confirmed a name or what it does. We’ll know a lot more by Wednesday’s report. Nothing to do yet, just watch for it.
So here’s your one thing for today. If you’re already on Muse, open it and literally ask it to get you into early access, then picture what a Mac app that runs tasks for you could take off your plate this week.
That’s today’s report. Transcript and every source linked at A I industry report dot com.
If you want your business to be the one AI recommends, the free checklist at the leveling window dot com is where everybody starts.
And full transparency, because we practice what we teach: this show is researched, written, and voiced by AI, built by our founder. We teach businesses what AI can do. This show is the proof. Hit follow, and tomorrow’s report finds you on its own.
Sources
- Meta opens early access program for new Muse features, via TechCrunch (Sarah Perez, September 25)
- Introducing the new Copilot with Home, Code and Autopilot, via Microsoft’s official blog (September 25)
- Microsoft Copilot overhaul rollout detail, via Unite.AI (September 25)
- Google Has Deployed a New AI Spam Detector Called SAFE, via Search Engine Journal (Roger Montti, September 25)
- Your AI content watermark isn’t what Google is looking for, via Search Engine Land (September 25)
- OpenAI prepares managed agents for DevDay 2026, via TestingCatalog (Alexey Shabanov, September 26; Rumor Watch, OpenAI has not confirmed)
- DevDay 2026 date and location, via OpenAI (background)
Want your business to be the one AI recommends? The free Agentic Readiness Checklist walks you through the 15 signals AI agents check, in plain English. Get it at thelevelingwindow.com.
The AI Industry Report is researched, written, and voiced by AI, built by the founder of the Agentic Marketing Academy. We teach businesses what AI can do. This show is the proof.
