Short answer
A company can often license its Slack or Microsoft Teams archive for AI training, because the organization generally controls its own workspace content. Chat logs also contain personal details and other companies' confidential information, so a deal needs a rights review, exclusions, de-identification and a written license. Chat is among the records buyers want most.
In April 2026, news reports described failed startups selling their old Slack messages to AI companies. By October, reporting described buyers offering six-figure sums to companies that were still operating. Team chat has become one of the most sought-after kinds of business data. Whether your company can license its own archive depends on four things covered here.
Key takeaways
- Chat archives are in demand because they record how people solve problems together.
- The organization, not the individual employee, generally controls a workspace's content. Check your own agreement with the platform.
- Chat is messy. It mixes work with personal details, HR matters and other companies' confidential information.
- A careful deal narrows the scope, excludes high-risk channels, de-identifies what remains and puts everything in a written license.
- How the archive is exported matters, because platform terms can restrict some methods.
Why AI data buyers want chat archives
A chat channel is a record of work in motion. Someone reports a problem. Colleagues suggest causes. One idea is tried and fails. Another works. That sequence is exactly what AI developers want their systems to learn, and it almost never appears in public text.
Ask whether the proposed use involves training a model, testing an agent or building a simulated workplace. Those are different uses; the license should say which are allowed. See why AI companies pay for business data.
Who controls a company's workspace data?
In business chat platforms, the customer organization generally owns or controls the content its workspace holds, and administrators can export it under the platform's rules. Employees write the messages, but they write them in a company system, usually under a policy that says so.
That gives the company a starting point, not a final answer. Control over a workspace does not override:
- promises the company made in its privacy policy, employee notices or handbook,
- confidentiality obligations owed to customers and partners whose information appears in the messages,
- privacy laws that apply to the people in the conversations.
What is inside a chat archive that needs care
| Content | Where it shows up | Usual treatment |
|---|---|---|
| Personal details | Direct messages, social channels | Exclude or de-identify |
| HR and personnel matters | Manager and HR channels | Exclude |
| Pay, scheduling and working-condition discussions | Team channels, direct messages | Exclude |
| Legal advice | Channels with counsel | Exclude, to protect privilege |
| Customer confidential information | Account and project channels | Exclude or de-identify, subject to contracts |
| Pricing logic and strategy | Leadership channels | Often exclude, to protect trade secrets |
A law firm that advises on these deals makes two points worth repeating. First, removing identifiers from millions of free-form messages is much harder than removing them from structured fields. Second, de-identification does not make content non-confidential. A complaint about pay is still sensitive after the name is removed. That is why high-risk channels are usually left out entirely.
Platform terms and how data is exported
Your agreement with the chat platform sets the rules for exporting content. Two points are worth checking with your attorney.
Export method. Platforms offer their own administrator export tools, with different options for public channels, private channels and direct messages depending on your plan.
API restrictions. According to a 2026 law firm analysis, Slack's API terms prohibit third-party app providers from using data obtained through the API to train a large language model. An archive pulled by a third-party app may be treated differently from one the company exports itself. Settle the method with the buyer in writing before anything moves.
It also helps to keep two questions apart. Whether a platform uses customer data to train its own AI is one question. Slack faced criticism over the wording of its privacy principles in May 2024. It responded that customer data is not used to train large language models, while some non-generative features, such as search and recommendations, learn from customer data unless an organization opts out. Whether your company chooses to license its own archive is a separate decision, and it is yours.
How a careful deal handles chat data
- Narrow the scope. Start with work channels: projects, operations, support. Many companies exclude direct messages altogether.
- Exclude high-risk channels outright. HR, legal, leadership, payroll and anything a customer contract restricts.
- Agree on de-identification. Who removes personal information, vendor names and identifiable customer details, with what method, and who checks the result. See de-identification for AI training data.
- Put it in a license. Permitted use, exclusivity, term and payment, in a written agreement you review and sign. See key terms in an AI data licensing agreement.
- Think about your people. Decide how and when employees will be told. See employee privacy when a company licenses workplace data.
Slack, Teams and Google Chat
The same reasoning applies to Microsoft Teams and Google Chat. Buyers care about the content and its history, not the brand of tool. What differs is the export tooling and the platform's terms, so check the agreement for whichever product you use. A company that moved from one chat tool to another may hold two archives, and both can count.
What DataRelai does and does not do
DataRelai is a consultant for companies that license their data to AI buyers. We qualify your data, find buyers and work with the buyer through closing. We never take or handle your raw data. There are no uploads and no access to your systems. Data moves only under a license agreement you sign with the buyer, and we only work with buyers who anonymize and de-identify data. See if your company qualifies.
Frequently asked questions
Do employees have to consent before a company licenses its Slack archive?
It depends on the law that applies and on what the company told employees. Reporting quotes a law professor saying no US federal law bans such sales, while arguing employers should get express consent. State privacy laws, labor law and the company's own notices can all apply. Ask your attorney.
Can we license only some channels?
Yes. Scope is negotiable. Limiting a deal to work channels and excluding direct messages is common.
Does Slack sell our messages to AI companies?
Slack said in 2024 that customer data is not used to train large language models. Platform terms can change, so check its current privacy principles. Licensing your archive is a decision your company makes, not the platform.
How much is a Slack archive worth?
It depends on team size, years of history, variety and the license terms. See how much company data is worth to AI companies.
Sources
- Frankfurt Kurnit Klein & Selz, "Selling Company Data for AI Training: What to Know Before You Say Yes" (October 1, 2026)
- The Register, coverage of Slack's AI privacy principles (May 20, 2024)
This guide is general information, not legal, tax or financial advice. Whether a particular dataset can be licensed depends on what it contains and on obligations your company already has.