Data brokers are actively purchasing internal corporate communications and workflow files to train artificial intelligence models, according to a report by Wired.reported The acquisition targets include emails, Slack messages, support tickets and call transcripts from sizable companies.Companies are selling their internal emails and messages to AI companies to help them learn how to do real work.Private enterprise data, including Slack and tickets, is being commoditized and sold to labs for model training at scale.
The figures
- $2 million: higher stakes offered in another pitch including call transcripts, per Wired.
- $45 billion: total spending by labs on private data, per a pitch cited by Wired.reported
- $1,000: offer made to a freelance writer for a single story, per Wired.
- 2023: founding year of industry leader Mercor, per Wired.
- 22: age of Mercor’s three cofounders when they became billionaires, per Wired.
Behind the figures
Wired reported that these figures come from unsolicited email pitches received by journalists and entrepreneurs.reported One sender, identified as Nick, claimed labs pay media companies for old work records. Another sender, Cora, stated that public data is exhausted. Handshake representatives confirmed they license professional documents for model training.
What it means for your stack
The data being acquired includes structured and unstructured internal communications such as Slack messages and support tickets. This suggests that proprietary workflow data is being ingested into model training sets. Developers should consider that internal company processes may now be part of the training corpus for major AI models.
Building on this
Developers can build applications that leverage models trained on real-world knowledge work, including journalism and graduate-level coursework. Interfaces can be designed to handle the nuances of professional documents that these models have been exposed to. Limits remain unclear regarding the extent of anonymization and the specific fields covered beyond journalism and student papers.
Why it matters
- Training data sources are shifting from public web content to private corporate workflows, altering the provenance of model outputs.
- A pitch cited by Wired claims labs are spending over $45 billion on private data, indicating significant economic activity in this sector.
- Internal company communications, such as Slack and emails, are now commodities being sold to labs, raising data governance concerns.
- Startups winding down may have their operational data sold to AI labs through brokers who scrub and anonymize the information.
Topics: data privacy, AI training, data brokers, corporate security
Updated
The strongest doubt
Counterpoint from the sources
Not everyone is thrilled with the acquisition of internal corporate files, as noted by Wired. A freelance writer declined a $1,000 offer to sell her work, indicating some resistance to selling personal professional output.
For readers
Developers
- Models may be trained on private corporate workflows, affecting output bias and data provenance.
- Internal communications like Slack messages are part of the training data landscape.
Policy
- The sale of private employee data raises significant privacy and labor law questions.
- Brokers claim to strip identifying details before selling data to labs.
Business
- Winding-down startups may have their operational data sold to AI labs.
- Companies may need to audit their data disposal processes to prevent unauthorized sales.
Numbers in context
Circle area is proportional to the magnitude; labels retain signs. Figures may describe different measures. Read their contexts above.Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Questions readers ask
Which AI companies are buying private data?
Pitches cited by Wired name Anthropic and OpenAI as labs paying for media company records.
What types of data are being purchased?
Buyers seek emails, Slack messages, support tickets, call transcripts, documents and graduate student coursework.
How are the data brokers handling privacy?
One broker, Sunset, stated they scrub and anonymize data, stripping identifying details before brokering deals.
Connected stories
Reading receipt
- AI labs reportedly pay $350K to $1.8M for single media company archives.
- Industry leaders include Mercor, Handshake and Micro1 in a booming sector.
- Brokers anonymize data from winding-down startups before selling to labs.