Data brokers are gathering companies' internal files to use as training material for AI models, and the practice is drawing criticism from some quarters.
AI training data · Data brokers
Data brokers are gathering companies' internal files to train AI models
The material is not publicly available, which makes it far more specialized than open-web data. The practice is drawing criticism, and key questions about consent, confidentiality and licensing remain unanswered.
How the data moves
A simplified flow based on the reporting
1. Companies
Internal corporate documents and files, not publicly available
→
2. Data brokers
Intermediaries collect and aggregate the files for resale
→
3. AI developers
Buyers seek specialized, high-quality data for training
→
4. Model training
Proprietary material used to make models more useful for specific tasks
What the reporting establishes
Brokers are collecting internal corporate files for AI training.
The files are not publicly available.
Demand for proprietary, task-specific data is growing.
Some affected parties have voiced discomfort with the arrangement.
What remains unclear
Which companies' files are involved.
Who the buyers are.
The terms under which the files were obtained.
The legal and contractual safeguards that apply.
Why it matters
Whether this practice is lawful depends on details not yet public.
Confidentiality
Do agreements allow these files to be shared?
Privacy
Do the files contain personal information covered by privacy rules?
Intellectual property
Who holds the rights to license the material?
Intermediary duties
What responsibilities do brokers carry for consent and governance?
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article...