Every morning, somewhere in a finance department, a tired analyst pastes a spreadsheet full of customer records into a chatbot and asks it to find the pattern. No malice. Just a deadline. That single paste is the problem Vinay Goel decided to build a company around.
Goel is the CEO and co-founder of Wald.ai, a Palo Alto startup that does something delightfully sneaky. When you ask an AI assistant a question, Wald.ai steps in the middle. It strips out the sensitive bits - names, account numbers, anything confidential - swaps in stand-ins, lets the AI answer the now-anonymous question, then quietly restores your real data before you ever read the reply. The chatbot gets a useful answer. The chatbot never gets your secrets.
He calls it contextual data protection. Most people would call it a magic trick. Redact, answer, restore - all of it inline, all of it encrypted end to end, fast enough that the person typing never feels the seam.
The insight underneath it is almost annoyingly obvious once Goel says it out loud. Employees are going to use AI whether you sanction it or not. The corporate firewall blocking ChatGPT is a speed bump, not a wall. So the honest question isn't whether AI shows up at work. It's whether you can see what happens when it does.
This is the part most security vendors get backwards. They reach for the off switch. Goel reaches for the dimmer. Bans, he argues, just push usage into the shadows where nobody can measure the risk. Visibility first, control second. Make the safe path the easy path, and people will take it without being nagged.