Slack AI safety profile

What it does with your data

Slack AI works on what is already in your Slack: messages, files and channels. It helps to separate two kinds of AI.

The first is Slack’s predictive models. These power emoji suggestions, channel recommendations, search results and autocomplete. By default, Slack analyzes customer messages, content and files to train these global models unless your organization opts out.

The second is the generative features, such as conversation summaries, recaps and AI search answers. Slack says it does not build generative AI models with customer data and will not do so without your organization’s affirmative opt-in. These features use third-party large language models. Slack says those models are not updated by your data and do not keep it after a request, and that it sends only the data needed for each task.

On storage, summaries and search answers disappear when you navigate away. Recaps are kept for 90 days and are deleted if the source messages are deleted. Channel summaries posted by the Summarize public channels workflow step follow your organization’s retention settings.

Settings to change first

Do these before staff start using AI features:

1. Opt out of global model training. An Org or Workspace Owner, or the Primary Owner, emails Slack’s Customer Experience team at feedback@slack.com. Include your Workspace/Org URL and use the subject line “Slack Global model opt-out request.” Slack replies once the opt-out is complete. After that, your data only improves your own workspace.
2. Choose which AI features members can use. Owners and admins control this. Turn off anything staff do not need.
3. Review AI search file settings. Admins can disable all file results or stop AI search from using externally hosted files.

The vendor pages checked do not give menu paths for these settings, so ask whoever manages your Slack workspace to find them.

Free and paid tiers

The pages checked state no plan difference for training or retention. The differences are about control and where data lives:

• Free and Pro: data is stored in the US. To choose a region, you must upgrade to Business+.
• Business+ and Enterprise: you can store certain data in one chosen region, such as Frankfurt, London, Montreal or Sydney, set up through Sales. Messages, files, canvases and the search index stay in that region. Member profiles, usage data and some other data may be stored elsewhere.
• Enterprise: admins can limit AI features to certain users and groups.

If you already use Slack and turn on a region later, older data stays in the US unless you ask for a migration.

Admin controls

On any plan (the pages state no limit), owners and admins decide which AI features members may use and can restrict file results in AI search. On Enterprise plans, Org Owners and Admins can also limit AI features to chosen users and groups. Only the Primary Owner can turn on data residency, on Business+ and Enterprise.

Slack’s compliance page lists ISO/IEC 27001, ISO/IEC 27017, ISO/IEC 27018, ISO/IEC 27701 and ISO/IEC 42001, plus SOC 2 (Type Ⅱ) and SOC 3. You can request the SOC 2 report through the Privacy and Security Form. Slack also says it can be configured for HIPAA compliance. These cover Slack in general, not the AI features specifically.

Using it with donor and client data

Any plan can work for a small non-profit, as long as an owner has sent the training opt-out email first. Choose Enterprise if you want only a few trained staff to have AI features, and Business+ or Enterprise if you need data stored outside the US.

The AI features read what is already in Slack, so the safest rule is to keep sensitive records out of Slack entirely. Do not post or paste:

• donor names with gift amounts
• payment or bank details
• client case notes
• health information

Get two things in writing. Keep Slack’s reply confirming your opt-out is complete. If anyone may store health information, get Slack’s HIPAA configuration terms in writing before they do. Also check that no one in your organization has given opt-in consent for generative AI training.

Questions people ask

Does Slack AI train on our messages?

Slack’s predictive models, like search results and emoji suggestions, train on your messages and files by default. Generative AI is not trained on customer data unless your organization opts in. An owner can opt out of the predictive training by email.

Where is our data stored?

On Free and Pro plans, data is stored in the US. Business+ and Enterprise plans can choose one region for messages, files and similar data. Slack says AI features keep customer data inside its own infrastructure.

How long are AI summaries kept?

Summaries and search answers vanish when you navigate away. Recaps are kept for 90 days. Summaries posted to a channel or canvas follow your retention settings.