AI's White-Collar Playbook: Contractors Upload Work for Training, Lawyers Warn of IP Time Bombs

OpenAI is building AI training data by asking contractors to upload actual work - but the process raises red flags for IP lawyers. The tech giant and Handshake AI are enlisting professionals to submit real work samples from past or current jobs for training datasets.

Contractors are instructed to delete proprietary information and personal identifiers using ChatGPT's 'Superstar Scrubbing' tool—a process that has drawn sharp criticism from legal experts.

Evan Brown, an intellectual property lawyer, said:

"Any AI lab taking this approach is 'putting itself at great risk'."

OpenAI declined to comment on the practice. Crucially, no details have been disclosed about how contractors are compensated, how long the data is retained, or what commercial usage terms apply to the uploaded content.

The absence of contractual clarity creates a legal gray area for both contributors and the AI firm.

The 'Superstar Scrubbing' tool, which relies on automated redaction, has known limitations in identifying sensitive patterns. Legal experts warn this could leave residual intellectual property in training datasets, potentially exposing OpenAI to litigation over unauthorized use of proprietary work.