OpenAI is reportedly asking third-party contractors to upload real work from their past and current jobs as part of a broader effort to gather high-quality training data for its AI models, according to a report by Wired. The initiative, which involves collaboration with training data company Handshake AI, aims to improve the accuracy and relevance of AI systems by incorporating authentic, real-world outputs. According to the report, OpenAI provided contractors with a presentation that instructs them to describe tasks they’ve performed in previous roles and upload actual work products. These materials can include documents such as Word files, PDFs, PowerPoint presentations, Excel spreadsheets, images, and code repositories—specifically emphasizing that the uploaded files should be concrete outputs, not summaries. Contractors are advised to remove any proprietary or personally identifiable information before uploading. OpenAI reportedly directs them to use a ChatGPT-powered tool called the “Superstar Scrubbing” feature to help sanitize the files. Despite these precautions, intellectual property lawyer Evan Brown warned that this approach carries significant legal and ethical risks. He noted that the strategy relies heavily on contractors’ judgment in determining what constitutes confidential or sensitive information, which could lead to unintentional leaks of trade secrets or private data. “This puts the company at great risk,” Brown said, highlighting the challenges of trusting external individuals to make such decisions. OpenAI has not publicly confirmed or denied the details of the program. A company spokesperson declined to comment on the matter. The move reflects a growing trend among AI companies to source training data directly from professionals across industries, particularly in white-collar fields like marketing, engineering, and finance. By using real work samples, these firms hope to train models that can eventually replicate or automate complex professional tasks. However, the practice raises concerns about data privacy, consent, and ownership—issues that could become increasingly prominent as AI systems grow more capable and widespread.