Apple Plans Siri Chatbot With Auto-Delete and Shorter Memory
Bloomberg’s Mark Gurman says Apple is preparing to make privacy the defining claim of its next Siri update, expected to be announced at WWDC, rather than competing only on chatbot capability. Gurman reports that the revamped assistant will let users automatically delete conversations after set periods and will retain less memory than many rivals, a trade-off Apple is likely to present as consistent with its long-running privacy pitch.
Apple is positioning Siri’s next phase around data retention, not just capability
Mark Gurman says Apple’s privacy pitch is not a new posture being invented for AI. It has been one of the company’s central product claims for “going over a decade,” and Apple routinely returns to it when introducing new hardware, operating systems, software, and services. In Gurman’s framing, privacy is Apple’s “North Star”: the company presents its products as built from the beginning around encryption, security, and reduced exposure of user data.
That positioning matters, he argues, because of what now lives on Apple devices: financial data, personal communications, smart-home information, and smart-home controls, among other sensitive activity. Gurman says Apple intends to extend that privacy posture into the new Siri experience, as it prepares to enter a market where “everyone’s got a chatbot” and AI features have become common across the major technology platforms.
The distinction Apple is expected to emphasize is not simply that Siri will answer questions or act more like a chatbot. Gurman says Apple will compare itself with companies such as Meta and Google by arguing that its chatbot retains less information and has less “memory” than competing systems. Bloomberg’s on-screen lower-third described the forthcoming product as a new digital assistant that may launch as a beta test; Gurman said Apple is expected to announce the chatbot at its developer conference on June 8.
The most concrete privacy feature is automatic deletion of chats
The specific feature Gurman reported is automatic deletion of chatbot conversations. Apple plans to let users decide how long chats remain available, using a model similar to the retention controls in Messages. A user could save chats on the phone and in iCloud forever, set them to delete after a year, or set them to delete after 30 days.
That is different, in Gurman’s description, from the “incognito mode” available in some other chatbots. Those modes resemble private browsing: a temporary private-chatting state. Apple’s proposed feature is a standing retention setting for conversations, not merely a special mode a user enters for a particular exchange.
The auto-delete of conversations means that that conversation, that chat, that's just gone, forever.
Gurman separates this from chatbot memory. Deleting a conversation removes the chat itself. Memory is a related but distinct system: it concerns what the assistant preserves from prior interactions in order to personalize or contextualize future responses. Caroline Hyde presses on the practical implication: if a user chooses 30-day auto-delete, what does that mean for the assistant’s ability to remember prior discussions and remain useful over time?
Gurman’s answer is that Apple appears to be trading off long-term personalization against privacy. He says some chatbots keep memory indefinitely, while others keep it only for a finite period. Apple, by contrast, is expected to retain memory for a shorter amount of time than much of the competition. That could make the product “useful in the short term” but “less useful in the long term.”
A shorter memory is the product trade-off
The tension is straightforward: a chatbot that remembers more can carry forward more context, while a chatbot that remembers less may fit Apple’s privacy positioning at the cost of durable personalization.
Mark Gurman says Apple is using a technique called differential privacy, which he describes as a way to retain some memory without having it “glued exactly” to the user. His description links that privacy approach to memory that can be stored on device and later used to provide context or tell the user what has been discussed, while avoiding a tighter association between stored information and the individual user.
That leaves Apple’s intended balance as the main product claim: preserve enough context to remain useful, while marketing the assistant around retaining less information than rivals and limiting how closely that memory is tied to the user. The company is not expected, in Gurman’s account, to emphasize the longest memory or the most persistent profile of the user. It is expected to argue that shorter memory, automatic deletion, and privacy-preserving handling of stored context fit the company’s broader privacy-and-security pitch.
I think it is going to work fairly well. It is going to be fairly competitive, but Apple's really going to market it around privacy and security.
The substantive distinction discussed by Hyde and Gurman was the privacy approach around Siri’s chatbot-style experience: less retained information, shorter memory, and user controls for how long conversations remain.