San Francisco-based Handshake AI is recruiting “Professional Document Contributors,” promising to pay up to $30,000 for “high-quality written documents” (approximately RMB 203,000 at the current exchange rate).
To apply for the position, candidates should have work experience in fields such as consulting, finance, law, software engineering, or data science. Handshake AI says applicants must own the relevant documents and be authorized to share them.
This has also raised a series of compliance and privacy questions: Which work-related documents truly belong to you and can be shared? And how much private work-related information should an AI have access to?
Handshake AI has listed a price for the documents: $6 per page (approximately RMB 40.5 at the current exchange rate).
Candidates whose submissions pass review will be paid at a rate of $6 per page (approximately RMB 40.5 at the current exchange rate). Each person may submit up to 50 documents, with a maximum of 100 pages per document. This means the maximum compensation for the project is $30,000 (approximately RMB 203,000 at the current exchange rate).
It is important to note that only pages that “pass review” will be paid for. However, the job listing does not clearly explain what criteria a document must meet to pass review.
Slides and presentations are “typically not eligible.” Handshake AI is looking for written materials, such as Word documents and PDFs.
The job listing states: “This is not a traditional full-time job. You only need to submit your information once. If you qualify, we will invite you to participate in document contribution projects when relevant opportunities arise.”
AI training companies are now paying for various types of datasets. Some people can earn money by helping fold clothes, while others can let free cleaners into their apartments as long as they allow the entire process to be filmed to provide data for training AI robots.
However, these companies often establish very strict rules, which may also limit the compensation ultimately available.
If you want to submit your own documents, you must first own the rights to those documents.
Thomas Ahlering, a partner at King & Spalding who specializes in data privacy, expressed concerns about the arrangement. First, he wants to know what the data will ultimately be used to train. “It is currently unclear whether the data will be used for their own models or transferred to other organizations,” he said.
He is also concerned about potential copyright issues. “Who actually owns these documents?” he asked. “What review mechanisms are in place to ensure that the documents contain no confidential information?”
A reporter asked Handshake AI how it would verify that applicants owned the relevant documents, but a company representative did not respond.
Proving document ownership is not easy. Some situations clearly would not qualify, such as work products you created for a company during your employment. Simply put, work products you think belong to you may actually belong to the company.
What if the document was created for a client? Or what if you created it yourself but it contains sensitive information from past business dealings? “I think this falls into a gray area,” Ahlering said.
For those hoping to make quick money from these documents, it may also be unwise to simply ask their employers to transfer the copyright to them. “I think most employers would be uncomfortable with that,” Ahlering said.