AI assistants want your memory
A new YouTube analysis says the next big fight in AI is over which assistant stores your long-term history — not just which one answers questions best. The video 'Anthropic And OpenAI Are Fighting Over Your Memory. You're Going To Lose.' frames memory as the feature that raises switching costs, expands privacy risk, and cements platform control (youtube.com).
AI companies are racing to make chatbots remember you, turning memory into the feature that can keep users locked into one assistant. (openai.com) OpenAI said on April 10, 2025 that ChatGPT memory became “more comprehensive” for paid users by referencing past conversations, and on June 3, 2025 it said free users were getting a lighter version with short-term continuity across chats. (openai.com) OpenAI’s help center says ChatGPT memory now has two parts: “saved memories” that users ask it to keep, and “chat history” that lets it draw on earlier conversations for future replies. The company says users can ask what it remembers, delete items, clear all memories, or turn the feature off. (help.openai.com) Anthropic rolled out memory for Claude Team in September 2025, saying Claude can remember a team’s projects and preferences across conversations with “project-specific boundaries” and user control over what is stored. (anthropic.com) The underlying idea is simple: a chatbot works better when it keeps a notebook on you. OpenAI says memory can retain details like your diet, goals, and recurring topics so you do not have to repeat them in every new chat. (help.openai.com) That notebook also raises the cost of leaving. If one assistant has months of your preferences, work habits, and ongoing projects, switching to another service means rebuilding that history from scratch, even if the rival model answers just as well. (openai.com) The privacy tradeoff is built into the product design. OpenAI says Temporary Chat does not use or update memory, does not appear in chat history, and is not used to train models, while regular chats can be used for model improvement if a user leaves training on. (openai.com; help.openai.com; help.openai.com) Anthropic has made a similar pitch for Claude, saying “Incognito chat” gives users a clean slate for conversations they do not want preserved in memory. Its memory product is narrower than ChatGPT’s public rollout, but it is aimed at the same habit: keeping useful context attached to the assistant instead of the user starting over each time. (anthropic.com) OpenAI also says advertisers do not get access to chats, chat history, memories, or personal details, even as the company adds ad controls to ChatGPT. That leaves the central contest less about banner ads than about which assistant becomes the default place where a user’s digital history accumulates. (openai.com) A recent YouTube analysis packages that fight as a battle over “memory” rather than raw model quality, but the product launches are already moving in that direction. The more these assistants remember, the less they look like one-off tools and the more they resemble accounts people may be reluctant to abandon. (youtube.com; openai.com; anthropic.com)
