Your deal advisor builds a data room for the sale. Contracts, financials, staff lists, systems. This guide covers one more folder most rooms leave out: your old records, and what AI labs pay to use them.
What AI data licensing is
AI data licensing means your company lets an AI lab use its records for a set time, under written rules. The records are things like Slack threads, email, CRM notes and support tickets. You keep ownership of the records.
In plain words
Licensing means a lab uses your records for a while, under rules.
Think of it like this
The lab rents the room. You keep the house.
Why would a lab want them? AI labs use real work to teach software agents how work gets done.
Epoch AI, a research group, reports that labs spend heavily on simulated workplaces for agents. It also describes lab partnerships with product companies and does not discuss real company archives. That gap is why buyer demand is still being tested, deal by deal.
In plain words
Labs want real work to learn from.
Think of it like this
It is a library that lends out real office life. Borrowed, not kept.
What public deals show
Public prices are few. We track every one we can find. Here is what the record shows today.
- A small shutdown range. SimpleClosure sold archives from nearly 100 small companies, at $10,000 to $100,000 per company, according to Gizmodo.
- One long history. Cielo24 sold 13 years of Slack, Jira and email for "hundreds of thousands" of dollars, per the same Gizmodo report.
- A big airline archive. Google won the auction for Spirit Airlines' archive with a $10 million bid, and court approval is pending. It held about 100 million emails, 500 million Teams chats and 30 million lines of code, per Time.
- Rival bids. Mercor bid $7.5 million. micro1 later offered $12.5 million, per Fortune.
- micro1's stated tiers. micro1 lists $100,000 and up for qualified companies, $500,000 and up for large ones, and $1 million and up for highly unique data, per its page. That is micro1's own claim.
Read the Spirit number with care. It was a bankruptcy sale of one airline's archive. Most companies sit far lower. See our deals page for each case with its source.
What can go wrong
Records hold other people's information. Daniel Goldberg, a partner at Frankfurt Kurnit, lists the main risks in a 2026 note.
Other companies' secrets sit in email under NDAs. Trade secrets and attorney-client privilege can be lost. Employee chat may be protected under labor law.
Three more points matter in a licensing deal.
- Privacy rules. Under California law, licensing personal data can count as a sale unless names and details are stripped out, per Freshfields.
- Vendor terms. Salesforce changed its Slack data-access terms in 2025 to limit bulk export and AI use of data pulled from Slack, per Hunton.
- Old promises. The FTC says quietly changing terms to allow AI training can be unfair or deceptive.
Our risks page has the full list in plain words. We do not give legal advice. Your lawyer decides what can be shared.
What your deal advisor should check
Your advisor does not need a long test. Ask them to look for a few plain signs in your company.
- Your company has 30 or more staff. micro1 sets that floor for its program, per micro1's partnership page.
- Your team has used Slack, email, a CRM or a ticket system for years.
- The records are in English and mostly work talk.
- You can say who holds the admin login.
- Customer contracts do not bar sharing, or someone can check.
If your company ticks these boxes, it is worth a quick check. It takes about three minutes on our data check, and you can bring the result to your advisor.
Where this fits in the sale
The listing is when the records come out of the attic. The data room is where your advisor lists them. The LOI is where you say who gets them. Our next guide covers the data line step by step.
Start with the free data check. If you want to talk it through first, send Matt a note. To see who buys what, read the buyer comparison.
