Recent reports of Microsoft Copilot accessing and summarizing confidential email messages have brought renewed attention to a fundamental question about AI: If a model has processed sensitive data, can it be told to forget it?
The answer depends on what we mean by “memory.” Understanding how AI models actually store
Register for free to keep reading
To continue reading this article and unlock full access to GRIP, register now. You’ll enjoy free access to all content until our subscription service launches in early 2026.
- Unlimited access to industry insights
- Stay on top of key rules and regulatory changes with our Rules Navigator
- Ad-free experience with no distractions
- Regular podcasts from trusted external experts
- Fresh compliance and regulatory content every day