What does my AI tool remember about my firm and who can see it?
By Gary Dhanda · 11 November 2026 · 3 min read
Short answer
Each tool handles memory differently. OpenAI and Anthropic say users can see and delete what is saved, and that organisation owners or admins on business plans can switch memory off. Google says Gemini in Workspace keeps conversation history and that admins set how long it is kept. None of the pages we read says whether an admin can read an individual's saved memories, so ask your provider before you rely on it.
ChatGPT
OpenAI's page on memory says you can see what ChatGPT has saved in Settings, under Personalization and Manage Memory, and that you can ask it what it remembers. It says deleting a chat does not erase the memories from it. You have to delete the memory itself.
You can turn memory off, and a temporary chat does not use it. For Enterprise, OpenAI says account owners can turn memory off for the whole organisation and that memories and other workspace information are excluded from training its models.
That page is dated 13 February 2024, with updates to June 2025. Check the current version before you rely on any detail.
Claude
Anthropic's help centre says Claude saves memories as you chat unless you switch the setting off. Its announcement of memory for Team and Enterprise plans, dated 11 September 2025, says each project has its own memory, that Enterprise admins can turn memory off for the whole organisation and that users can export their memory.
Gemini in Workspace
Google's admin help page, last updated 26 October 2026, says conversation history lets users see and continue past chats with Gemini in Workspace. Whether users can delete their own conversations, or choose how long they are kept, depends on what the admin allows.
Admins can set automatic deletion of inactive conversations at 90 days, 540 days or 1080 days, or choose no automatic deletion. The page does not describe memory in the sense the other two tools use the word, so we have only covered history here.
Who can see it
The pages we read describe who can switch memory off, delete it or set how long it is kept. None of them says whether an admin can read an individual's saved memories. We have not found that stated, so do not assume either way.
Ask your provider or check your admin settings, and write down the answer with the date.
What to do this week
List the AI tools your team uses and the plan each one is on. Check whether memory is on and who can change it. Look at what has been saved and remove client personal details. Our six checks set this out on one page, and our guide on client data in AI tools covers what to keep out.
Where this leaves you
The controls exist but they sit in different places and have different names. Someone in the firm needs to own the list.
This is a general guide, not legal advice. If you would like us to go through your tools with you, get in touch through our Contact page.
Sources
OpenAI, 'Memory and new controls for ChatGPT', openai.com/index/memory-and-new-controls-for-chatgpt, 13 February 2024 with later updates. Read 11 November 2026.
Anthropic, 'Bringing memory to Claude', claude.com/resources/articles/memory, 11 September 2025. Read 10 November 2026.
Anthropic help centre, article 11817273 on Claude's memory, support.claude.com/en/articles/11817273. Read 10 November 2026.
Google Workspace Admin Help, 'Manage Gemini in Workspace conversation history settings', knowledge.workspace.google.com, last updated 26 October 2026. Read 11 November 2026.