Anthropic has declared today that every paid Claude user will now possess memory. Earlier, in September, the AI firm disclosed that Claude memory was accessible to select paid users, particularly those subscribed to Team and Enterprise plans. Now, individual Max and Pro users will also gain from Claude’s capacity to “remember.”

Max subscribers can enable Memory in Settings right away, while Pro users will gain access to it in the upcoming days. Once the feature is activated, Claude starts constructing context from a baseline provided by the user. Users can transfer memory from ChatGPT or Gemini through copying and pasting, and they can also export memory from Claude.

Memory is optional and can be toggled on or off. Users can choose to erase specific memories or make use of Claude incognito, a feature recently rolled out by the company.

In a press release distributed to Mashable, Anthropic mentioned that it performed safety evaluations with memory to guarantee that Claude would not retrieve harmful discussions or become excessively compliant with potentially damaging requests, a concern noted in other LLMs like ChatGPT. Following these evaluations, modifications were implemented in how memory operates.

Upon releasing its latest lightweight model, Claude Haiku 4.5, Anthropic asserted it was the most secure version to date, reaffirming that its latest LLMs are safer than those that came before.

Concerning memory, the release emphasizes that Claude offers “complete transparency,” enabling users to view the “actual synthesis, not vague summaries” of what the model retains.

Anthropic has rolled out several enhancements to Claude in recent months, such as a new coding model, the compact Haiku 4.5 model, a Chrome extension, and the capability to generate spreadsheets and decks. Memory is backed by the Claude 4 model family.

“We’re advancing toward Claude comprehending your entire work context and adapting automatically,” remarked Anthropic chief product officer Mike Krieger in the press release. “Memory commences with project continuity, but fundamentally, it’s about establishing prolonged thinking partnerships that develop over weeks and months.”

Disclosure: Ziff Davis, the parent company of Mashable, initiated a lawsuit against OpenAI in April, claiming it violated Ziff Davis copyrights in the training and operation of its AI systems.