> At a Glance
> – OpenAI and Handshake AI ask contractors to upload actual past job files
> – Contractors told to strip proprietary data before sharing
> – Move aims to build training data that can automate white-collar tasks
> – Why it matters: Your workplace documents could help train AI that might one day replace similar jobs
OpenAI and its partner Handshake AI have launched a data-collection push that relies on contractors sharing genuine work they’ve completed for past or current employers, according to a Wired report. The effort targets high-quality training material that could push AI models deeper into office workflows.
How the Program Works
Contractors receive a presentation that walks them through describing tasks they’ve handled on the job. They’re then prompted to upload the actual deliverables-Word documents, PDFs, PowerPoint decks, Excel sheets, images, or code repositories-rather than summaries.
The guidelines tell participants to:
- Remove proprietary content
- Strip any personally identifiable details
- Use the ChatGPT “Superstar Scrubbing” tool for redaction
Legal Risk Spotlight
Intellectual-property attorney Evan Brown warned Wired that the approach puts “great risk” on any AI lab that adopts it, because success hinges on “a lot of trust in its contractors to decide what is and isn’t confidential.”
An OpenAI spokesperson declined to comment.
Industry Context
The request fits a wider trend: AI companies are paying outside contractors to create or contribute real-world data sets. The goal is to train systems that can eventually shoulder knowledge-work tasks now done by humans.
Key Takeaways
- Contractors must judge for themselves what is safe to share
- Mistakes could expose sensitive corporate or personal data
- The initiative underlines Big Tech’s appetite for office-grade training examples
With AI labs racing to automate knowledge work, every uploaded file moves the industry one step closer to models that can mimic-and potentially replace-human professionals.
