As competition in generative AI intensifies, Anthropic announced on September 12 that its Claude AI model is introducing a "memory" feature designed to help users better manage projects and preferences, but it is limited to paid Team and Enterprise plans. This update follows Claude's recent additions for creating spreadsheets and presentations, further strengthening its position in the enterprise AI tools market.
Memory details: an enterprise‑grade "smart assistant"
The memory feature is designed for work scenarios and can retain professional context such as client requirements, project details, and team workflows. When a user enables the "project" feature (a self‑contained workspace), Claude creates separate memories for each project to avoid information mixing. For example, when handling a marketing campaign it can automatically recall last week’s brand guidelines to produce consistent copy suggestions. This is more efficient than traditional stateless AI conversations and is expected to improve team collaboration by 20%–30%.
Activation is simple: enable it in the Claude App settings to view what Claude remembers and adjust the "memory summary" — users can choose to focus on or ignore specific details, such as excluding temporary meeting notes. Enterprise administrators have broader controls and can disable memory for an entire plan at any time to ensure compliance. Anthropic notes that users can import memories from other AIs (such as ChatGPT) or export Claude data, with documentation available at https://support.anthropic.com/en/articles/12123587‑importing‑and‑exporting‑your‑memory‑from‑claude。 Free users and individual Pro/Max plans do not yet have this feature; Anthropic says it will consider phased expansion through evaluation and testing.
This capability is not unprecedented: ChatGPT has supported personalized memories since 2023, and Gemini extended multi‑device sync in 2024. Anthropic’s implementation places greater emphasis on enterprise privacy; the company wrote on its blog, "Memory introduces new security considerations and is designed for work to avoid archiving sensitive conversations." While specifics were not detailed, this likely includes encrypted storage and automatic filtering mechanisms to align with requirements such as the EU AI Act.
Incognito chats: a privacy option for everyone
To balance the convenience of memory with privacy concerns, Anthropic simultaneously launched incognito chats available to all users. These conversations do not appear in history and are not saved to memory, making them suitable for brainstorming sensitive ideas or running experiments. Similar to a browser’s incognito window, users can switch to it at any time to ensure "no‑trace" interactions. An Anthropic product lead wrote on the blog, "Incognito chats make Claude more flexible, whether you’re a free user or an enterprise customer." In the era of GDPR and CCPA, this feature could attract privacy‑sensitive users.
Background and impact: Anthropic’s catch‑up strategy
For enterprise customers, the change signals Claude’s evolution from a "chatbot" into a "project partner," potentially saving teams up to 30% of their time. However, free users have complained of a two‑tiered system: posts under the #ClaudeMemory tag on X exceeded 5,000, calling for wider access. Analysts predict that memory features will become standard in AI by 2026 and could drive market growth by about 15%.
Claude’s memory update is a strategic move by Anthropic to balance convenience and security and to support enterprise AI transformation. Making incognito chats available to everyone underscores a user‑first approach. Anthropic concluded its blog post saying, "We will continue testing to ensure memory is reliable." In a market led by ChatGPT, this feature could help Claude make strategic gains, and the tech community is watching closely.
Anthropic’s decision also reflects a broader industry trade‑off: build AI that "remembers you" more effectively, or maintain short‑lived, controllable conversations. Anthropic has opted for a phased roll‑out backed by incognito and administrative controls to improve user experience while keeping risks within monitorable bounds. In the coming months, the company’s pilot data and user feedback will be key to refining the memory design.