Workplace chat data for AI training
Channel and direct-message history with threads, reactions, timestamps and file references from Slack, Microsoft Teams, Google Chat and similar tools.
- Measured in
- Messages, channels, active users and months of history.
- Who buys
- Labs and enterprise AI vendors building workplace assistants and agents.
- Where it comes from
- Operating companies, closed startups and businesses being wound down.
- Rights to clear
- Platform terms of service and API terms, employee handbooks and notices, and confidentiality obligations to clients.
- Privacy
- Direct identifiers, personal conversations and HR or health discussions removed. Direct messages often excluded entirely.
- Pricing signal
- No public rate card. Part of larger workspace bundles in most reported deals.
- Workplace chat from Slack or Microsoft Teams shows fast, informal coordination that email does not capture.
- Slack's API terms restrict using data obtained through the API to train large language models, so how the data was exported matters.
- The Spirit Airlines data Google bought in August 2026 reportedly included about 500 million Microsoft Teams messages.
- Workplace chat needs heavier de-identification than most text because people share personal details casually.
Why workplace chat is in demand
Agents that work inside a company need to understand how teams talk: short messages, shared shorthand, links to tickets and documents, and decisions made across a thread. Workplace chat is the clearest record of that behavior, and it is rarely available on the open web.
Platform terms come first
How the data leaves the platform matters. Slack's API terms include restrictions on using data accessed through the API to train large language models (Slack API terms). Admin exports, compliance exports and API access can be governed by different terms, so the export route needs to be checked before a buyer is told the data is available. Microsoft 365 data held by the company is generally governed by the company's own policies and agreements, but the same employee and client questions apply.
Employee and client questions
- Employee notice. What did your handbook and acceptable use policy say about monitoring and use of workplace messages?
- Protected activity. Discussions about pay, scheduling or conditions may be protected under the National Labor Relations Act.
- Client confidentiality. Channels shared with clients, or channels discussing client work, are normally excluded.
Preparing a workplace chat dataset
Most sellers exclude direct messages, private channels and anything shared with external organizations, then license public channels after de-identification. That keeps the parts with the most training value (project coordination) while dropping the parts with the most risk.
Deal context
MIT Technology Review reported that labs and data vendors are buying workspace data, including Slack and Teams history, from shut-down startups, with the largest reported example being Spirit Airlines (MIT Technology Review). If you are closing a company, read selling data from a closed company.
Questions
Can I sell my company's Slack history for AI training?
Possibly, but check how the data would be exported first. Slack's API terms restrict LLM training on data accessed through the API. You also need to clear employee notices and client confidentiality, and de-identify the messages.
Should direct messages be included?
Usually not. Direct messages carry the highest privacy risk and the lowest consent clarity. Most deals license public channels only.
Do buyers want files shared in chat?
Sometimes, but files are reviewed separately because they often contain client documents and confidential data.
Related guides
Related data types
Sources
Need workplace chat data, or hold it?
Tell us what you need or what you hold. We check rights and privacy first, then introduce vetted counterparties. You license directly with each other.
- We reply to every enquiry within two working days.
- No upfront fees for sellers
- We never hold your raw data