Anthropic has just made a bold move in the AI assistant wars. On March 1, 2026, the company launched Import Memory, a feature that could redefine the rules of competition: in under 60 seconds, users of ChatGPT, Gemini, or any other chatbot can transfer their entire memory data to Claude. A direct attack on OpenAI's most powerful lock-in mechanism.
Why this is a turning point for the industry
A chatbot's memory is its secret weapon for user retention. The more you chat with ChatGPT, the better it understands your preferences, writing style, and ongoing projects. This isn't just data — it's your digital fingerprint, reflecting how you think and work.
Until now, leaving ChatGPT meant starting from scratch. Months of mutual learning, gone. This switching cost mechanism is exactly what Anthropic just shattered with this feature.
How memory import works
The process is deliberately minimalist. No files to export, no JSON to parse, no API tokens to configure. A simple copy-paste is all it takes:
- Visit claude.com/import-memory
- Copy the migration prompt provided by Anthropic
- Paste it into your current chatbot (ChatGPT, Gemini, Copilot…)
- The chatbot generates a text block containing all its stored memories about you
- Copy that block and paste it into Claude's memory settings
- Claude automatically processes the information into its own memory system
An explosive context: the OpenAI user exodus
This announcement is no coincidence. Recent weeks have been turbulent for OpenAI: ads introduced in ChatGPT for free users, the permanent retirement of the beloved GPT-4o model, and most notably, a controversial partnership with the [Pentagon](https://www.defense.gov) for military use of its AI models.
Meanwhile, an OpenAI researcher resigned and published a New York Times op-ed titled "OpenAI Is Making the Mistakes Facebook Made." The result: Claude climbed to #1 on the App Store in the US, Germany, and Canada, overtaking ChatGPT for the first time.
“Competition among AI models has expanded from performance benchmarks to ecosystems and data sovereignty. User data portability is likely to become the key variable in the next phase.”
What about privacy?
Anthropic emphasizes the security of imported data. Here are the stated guarantees:
- Imported memories are end-to-end encrypted
- They are never used for model training
- They are exportable at any time — no new lock-in
- One-click deletion available from settings
Google is already preparing its response
Google is currently testing a similar feature called "Import AI Chats" for Gemini. Key difference: Google's tool only transfers conversation logs, not memories. An important distinction, as memories represent a much richer contextual synthesis than simple chat logs.
What this means for the future of AI
Anthropic's initiative marks the beginning of a new era: AI data portability. Just as GDPR mandated personal data portability in Europe, the question now extends to AI interaction data. Your memories, preferences, and context — they all belong to you.
If this trend continues, AI assistants will no longer be able to rely on lock-in to retain users. They'll have to differentiate through the pure quality of their service. And that might be the best news for users.
Summary: AI data portability comparison
| Feature | Claude | ChatGPT | Gemini |
|---|---|---|---|
| Memory import | ✅ Yes | ❌ No | 🔄 Testing (logs only) |
| Memory export | ✅ Yes | ✅ Yes | ❌ No |
| Contextual memory | ✅ Yes | ✅ Yes | ⚠️ Limited |
| Price (pro plan) | $20/month | $20/month | $20/month |
| No training data use | ✅ Guaranteed | ⚠️ Depends on settings | ⚠️ Depends on settings |
Compare these AI assistants
Sources and references
Official websites and resources :
- Anthropic — anthropic.com
- ChatGPT — chat.openai.com
- Claude — claude.ai
- OpenAI — openai.com
- Pentagon — defense.gov
- Google — google.com
- GDPR — gdpr.eu
See our detailed reviews :
Stay updated on AI news
Get the latest AI tool updates delivered straight to your inbox.
No spam. Unsubscribe in 1 click.
