How to find and export your MySQL data for AI training

01

What is MySQL data, and why do AI labs and data vendors buy it?

What a MySQL database records

MySQL is a free, widely used database system, the software that stores a company's records and answers questions about them. A MySQL server holds one or more schemas (MySQL's word for a database), and each schema holds tables, which are grids where every row is one record, such as one invoice, and every column is one fact about it, such as the amount.

MySQL sits behind a large share of websites and online software, so its tables often contain a company's full transaction history. A row in an invoices table points to the customer row it belongs to by a number called a key, and every row carries the time it was created.

Why a linked transaction history teaches an AI model something

An AI model learns by studying a very large number of examples and finding the patterns in them, and those examples are called training data. A table of invoices tied to customers and support events shows what happened, in what order and with what result. Text from the public internet rarely contains records of that kind.

AI labs now want to train AI agents, software that carries out a job step by step, and an agent learns a job from records of that job being done. A data vendor can turn anonymized histories into worked examples, or into a practice copy of a business process in which an agent is scored on each attempt. The full mechanism is in AI training data: why AI companies want your company's data.

How MySQL fits into a dataset

A MySQL database is usually one of several sources in a company's dataset, next to email, documents and tickets. Companies are paid $100K to $2M or more per dataset, according to micro1, the data vendor we typically refer companies to. The price covers the whole dataset, not each source on its own, and what your data is worth explains what moves it.

02

Where does MySQL show how much data you have?

How to read your database size

The steps below are run by a developer or database administrator in the mysql command-line client or any SQL tool. They rely on the columns documented in the MySQL 8.4 manual page for the TABLES view.

1. Connect with an account that has SELECT access to every application schema, so that no table is missed.
2. Run this query. It adds the data and index sizes of each schema, converts bytes to GB by dividing by 1024 three times, and skips MySQL's own system schemas:

   SELECT table_schema,
          ROUND(SUM(data_length + index_length) / POW(1024,3), 2) AS size_gb,
          SUM(table_rows) AS approx_rows
   FROM information_schema.TABLES
   WHERE table_schema NOT IN ('mysql','information_schema','performance_schema','sys')
   GROUP BY table_schema
   ORDER BY size_gb DESC;

3. Ask for fresh numbers if the data changed recently. MySQL caches these statistics for 24 hours by default. Running SET SESSION information_schema_stats_expiry = 0; before the query turns the cache off for your session, and ANALYZE TABLE on your largest tables refreshes their figures.
4. Count rows exactly where it matters. For InnoDB tables (MySQL's standard table type), TABLE_ROWS is an estimate that can be off by 40 to 50 percent, so run SELECT COUNT(*) on your main tables.
5. Find the earliest created date in your core tables, such as the oldest row of an orders table, to learn how many years of history you hold.

What the size columns mean

The TABLES view reports sizes in bytes, and each column describes a different part of a table.

Why the size and the row count are estimates

These figures describe allocated pages, which are larger than the raw rows, and they come from cached statistics. The size to report is the sum of DATA_LENGTH and INDEX_LENGTH for the schemas you plan to share, and the row counts are a guide to scale, not an audit.

03

What should you know about your MySQL data before you apply?

The size in GB

Write down the summed DATA_LENGTH plus INDEX_LENGTH in GB from the query above. Because InnoDB reports allocated pages, the number comes out larger than the raw rows and larger than a dump file. Count the primary server only, not replicas and not binary logs (the running record of changes MySQL keeps for recovery and replication).

The years of history

The years of history come from the earliest created timestamp in your core tables. A SELECT MIN(created_at) on a main table, with the table's own date column, gives the date.

The number of seats

For a database, the seat count is the number of application users or customer accounts that the records describe, taken from a users or accounts table. MySQL's own logins are not seats.

What you don't need to do first

You don't need to export anything to apply, and you don't need to clean or scrub the database. The data vendor anonymizes the records before anyone else sees them, and nothing moves until a lawyer-reviewed license is signed. The application, the intro call and each later step are laid out in how to sell your company's data.

04

How do you export your full MySQL history?

Run the dump against a replica

A replica is a second MySQL server that keeps a live copy of the first. Dumping from a replica, where you have one, keeps the reading work off the server your business depends on.

The steps with mysqldump

mysqldump ships with MySQL and is described in the MySQL 8.4 reference. The person running it needs an account with SELECT on the dumped tables, SHOW VIEW for views and TRIGGER for triggers.

1. Take a consistent dump of an InnoDB database without locking its tables, including its stored routines, triggers and scheduled events:

   mysqldump --single-transaction --routines --triggers --events --databases app_db > app_db.sql

   Change app_db to the name of your schema, and list several names after --databases to dump more than one.
2. Keep the file for the transfer. It contains SQL statements, so reading it back replays every row, which is slow for a large database.

The steps with MySQL Shell, for large databases

MySQL Shell is a separate command-line program, and its dump utilities are described in the MySQL Shell 8.4 manual. The account needs the EVENT, RELOAD, SELECT, SHOW VIEW and TRIGGER privileges, and the output folder must be empty or not yet exist.

1. Open MySQL Shell and connect to the server.
2. To dump the whole instance, run:

   util.dumpInstance('/backup/full-dump')

3. To dump chosen schemas only, run:

   util.dumpSchemas(['app_db'], '/backup/app-dump')

How the two tools compare

Both tools copy the same records, and they differ in speed, format and how well they handle size.

What a dump includes and leaves out

A dump holds the rows, the definitions of the indexes, and the routines, triggers and events you asked for. It does not hold the index contents, binary logs or replicas, so its file is smaller than the size you measured. Tables that use a storage engine other than InnoDB are not covered by the consistency guarantee, which matters if your schema still has older MyISAM tables.

How long an export takes

MySQL gives no time estimate, and its own manual warns that mysqldump is not meant as a fast or scalable way to back up large amounts of data. Parallel dumps from MySQL Shell finish sooner on a big database, so start early and prefer a replica.

Sources: MySQL documentation, The INFORMATION_SCHEMA TABLES table, mysqldump and MySQL Shell instance and schema dump utilities. Steps checked against these pages in November 2026.

05

What stays private when MySQL data is licensed?

You choose what is in scope

A dump covers only the application databases you name, so the mysql system schema, which holds user accounts, and any tables of payment details stay out of the export. The data vendor also deletes records too sensitive to clean safely instead of anonymizing them.

Names are replaced before a lab sees them

Anonymization means replacing every detail that identifies a person or a company, such as a name or an email address, with a stand-in that is the same every time it appears. The rows still link to each other and the dates still run in order, so a lab can learn from them, but they no longer say who anyone is. The diagram shows one invoice before and after.

The method, including what is deleted outright and how you review samples, is in how company data is anonymized for AI training.

Ownership and payment

Licensing gives the buyer limited rights to a copy, while the company keeps ownership of the originals. The terms that matter, such as exclusivity, permitted use and deletion, are covered in data licensing. micro1 pays in full within 30 days of accepting the data, usually as a single payment.