Your AI Is Getting a Moving Truck (And Some Platforms Are Changing the Locks)

Four companies made announcements this week that all circle the same question, and none of them seem to realize the others exist. Who owns the relationship between you and your AI?

Google and Apple said “you do.” Wikipedia and Reddit said “yeah, but humans still run things around here.”

Both answers matter if you write books for a living.

The Great AI Migration

Google rolled out Import Memory and Import Chat History for Gemini this week. The pitch is simple. You copy a prompt from Gemini, paste it into whatever AI you’ve been using, and paste the response back. Gemini now knows what your old AI knew about you. Your preferences, your projects, your bizarre insistence on Oxford commas in dialogue tags, that one very specific feedback style you demanded at 2 AM. There’s also a chat history import where you can upload a full .zip of your old conversations (up to 5GB) and pick up right where you left off.

Anthropic did something similar with Claude a couple weeks ago.

Meanwhile, Bloomberg reports that Apple’s iOS 27 will let you choose which AI chatbot powers Siri. An “Extensions” system will let you plug Claude, Gemini, or presumably others into your phone’s voice assistant. Siri becomes the front door, but you pick who’s actually home.

If you’ve spent months training an AI to understand your writing voice and your series bible, that context is no longer trapped in one app. That’s new. A year ago, switching AI assistants meant starting over. Explaining yourself from scratch. Re-teaching every preference. Now there’s a moving truck.

And on the Apple side, talking to Claude through Siri while you’re driving and dictating scene notes, without opening an app, without copy-pasting anything… that’s the kind of integration that turns an AI from a tool you visit into a tool that’s just there.

The Competitive Part

Google and Anthropic aren’t building import tools out of the goodness of their hearts. (Shocking, I know.) They’re building them because the AI assistant market is turning into a land grab, and the easiest way to win users is to remove the friction of leaving someone else.

Same playbook as phone number portability and bank account switching services. When you make it easy to leave, you’re betting that more people will come to you than leave you. It’s a confidence move.

For authors, this is entirely good news. These companies are fighting for your loyalty by making their products better and your exit cheaper. That’s the kind of competition that tends to work out well for the people doing the choosing. (That’s us. We’re the people. Hi.)

Meanwhile, On the Other Side of the Internet

While Google and Apple are rolling out the welcome mat for AI, Wikipedia and Reddit are reinforcing the walls.

Wikipedia formally banned AI-generated articles last week. Editors can still use LLMs for basic copy editing or translation (with restrictions), but you can’t use AI to write or substantially rewrite an article. The community had been dealing with a flood of AI slop for months and finally codified what most editors already believed. An encyclopedia written by volunteers loses something essential when the volunteers aren’t actually writing. Fair enough.

Reddit is going a different direction but with the same instinct. Accounts that exhibit “automated or otherwise fishy behavior” will need to verify that a human runs them. CEO Steve Huffman put it pretty simply. “We want to make sure that when you’re on Reddit, you know when you’re talking to a person and when you’re not.”

Neither of these is anti-AI in the way the headlines make it sound. Wikipedia still allows AI as a copy editing aid. Reddit isn’t banning bots, it’s requiring them to identify themselves. These are platforms saying AI can be here, but it can’t pretend to be something it isn’t. And yeah? That’s a line I’m comfortable with.

One More Thing

David Sacks is no longer the White House AI and Crypto Czar, having used up his 130-day stint as a special government employee. He’s moving to an advisory council that will “study issues” and “make recommendations,” which is Washington for “you still get a lanyard but not a desk.”

The most consequential AI policy role in the federal government is now effectively vacant. Whatever regulatory framework comes next for AI tools just lost its most aggressive champion. Keep an eye on that one.