Your company blocked ChatGPT for sensitive files. Grab the guide to strip the name, the address, and the price, and the block stops mattering.
"Don't Upload Sensitive Files to AI." Okay, well now what? The hard part is deciding what a model needs to know, and who should have to make that decision.
TL;DR
- Employees are stressed by conflicting company instructions regarding AI use and data privacy.
- The consequence of disappointing a manager often outweighs the perceived risk of privacy violations.
- Employees are increasingly tasked with inventing their own privacy processes for AI tools.
- The article highlights the dilemma faced by auditors trying to balance client data protection with AI efficiency.
- Simple rules like 'don't upload the file' do not solve the problem of how to complete the work.
- Practical solutions involve creating routing, tiers, and local pipelines rather than relying on memory.
- The effectiveness of a company's privacy system can be tested by comparing approved paths with consumer routes.