Meta's New AI Remembers You Across Chats. Find Out What Your Staff Has Told It.
Meta released two new models on Monday called Muse Wonder and Muse Wonder Full, and the feature they're pushing hardest is something they call Active Memory. It carries important personal context from one conversation to the next and pulls up the relevant pieces when a task needs them. Meta also says it plans to use the models in its consumer products, and the reporting I read says the bigger plan is for Muse to get deeper access to your messages, calendar, and private documents.
Meta's models and ChatGPT and Gemini and the rest are all heading this way. An AI that forgets you after every chat is annoying to use, so they're building ones that don't forget. Fine. That's useful. But I've been in enough small businesses to know what people type into these things when nobody's looking.
What's already in there
Think about the last month at your shop. Somebody pasted a customer's complaint email into a chatbot to get help with the reply. Somebody else dropped in a spreadsheet of quotes to clean up the formatting. Your office manager asked about a payroll question and used a real employee's name. None of that felt like a big deal at the time, because the chat closed and that was that.
With memory turned on, that's not how it works anymore. The tool can hold on to what it learned, and it brings it back later, in a different chat, maybe on a different task, maybe for a different person if the account is shared. I don't know exactly how long Meta keeps things or what controls you get, and the coverage I found didn't say either. Meta says a fuller safety report is coming soon. I'd want to read that before anybody at my company relies on it for client work.
The fifteen minute check
You don't have to wait on Meta to do this, because it applies to every AI tool your people use. 1: Ask your team, plainly and without making it a trap, which AI tools they use for work, including the free personal accounts on their own phones. You'll hear about two or three you didn't know about. 2: Open the settings on each one you control and find the memory setting. Most have an off switch, and most have a page where you can see what it's saved. Go read that page. It's usually a little eye opening. 3: Delete anything in there that has a customer's name, an address, a price you quoted, or anything about an employee. 4: Pick one rule and tell everybody. Mine is simple: no customer names, no employee info, and nothing you wouldn't want in an email that got forwarded by accident.
If you're on a shared login, fix that first. Memory on a shared account means everybody's context gets mixed together, and then one person's customer details can come up in another person's chat.
I'm not telling you to turn memory off everywhere. For your own marketing ideas and your own writing habits it's great, and it saves me real time. The line I draw is anything that belongs to somebody else. Their information isn't mine to hand over to a company's servers, and I'd rather find that out in a settings page than from a customer.
I run AI workshops and one-on-one AI consultations for companies around Toledo, Northwest Ohio, and Southeast Michigan. If you'd like help writing the rules for your team, or you want somebody to sit down and go through your AI tool settings with you, email me and we'll set something up.
Email Jayson