The AI party gets a privacy remix
If the last wave of AI was all about “move fast and upload everything,” this one is more like: maybe don’t hand the keys to the kingdom to a chatbot. Former OpenAI exec Mira Murati’s startup, Thinking Machines Lab, released Inkling, an open-weight model that companies can customize without exposing their training data or source code.
Why executives are suddenly twitchy
The timing is doing a lot of work here. Microsoft’s Satya Nadella and Palantir’s Alex Karp both spent the week warning that businesses risk giving away their intellectual property when they rely too heavily on centralized AI models. Nadella’s version: companies pay once in cash and again in the proprietary knowledge they have to reveal to make the tech useful.
Karp was even spicier, basically arguing that enterprises are worried they’ll pour their best ideas into the machine and get very little back except a bill and a data headache. In other words: the AI equivalent of paying for a gym membership and then letting the gym use your workout plan to open a competing franchise.
What investors should care about
This isn’t just philosophy class for founders. It’s a real market signal that the next battleground in AI may be control, not just model size.
- Companies may prefer open-weight or on-prem-style AI setups if they’re guarding trade secrets.
- That could help software and infrastructure vendors selling customization, security, and private deployment.
- It also pressures big frontier labs to prove they can offer enterprise value without turning customers into data donors.
Big picture
The AI story is shifting from “who has the smartest model?” to “who lets you keep your crown jewels?” And if Murati’s release catches on, the winners may be the companies that make AI feel less like a confessional and more like a locked vault with a very smart assistant inside.