Anthropic added persistent memory to its Claude Enterprise tier, enabling the model to retain context and learned patterns across separate user sessions. The feature is available immediately to Enterprise customers with granular controls over what information is retained, shared, or forgotten.
The implementation differs from OpenAI's approach announced earlier this week — Anthropic's system uses explicit memory management APIs that give developers fine-grained control over memory contents, while OpenAI's GPT-5.1 memory operates more autonomously.
Security-conscious enterprises may prefer Anthropic's explicit model, which provides auditability and clearer compliance with data governance requirements. However, it requires more integration work from development teams compared to OpenAI's automatic approach.
The bigger picture: Persistent memory is quickly becoming a table-stakes feature for enterprise AI platforms. Expect Google and Microsoft to follow with similar offerings within the quarter.