TAWS

From the Wednesday, November 4 show

OpenAI staff say security was weak for years

AI-generated from the transcript

What happened
Five people told Reuters about security problems at OpenAI going back years: too many workers could see too much code, and internal tools weren't being checked. Some staff raised alarms before ChatGPT launched in 2022. In October an engineer was arrested over a hidden recording device in an executive room; he is charged, not convicted.
The claim
OpenAI says protecting its tech is a matter of national importance.
Austin's take
Austin says your chats sit in data centers somewhere, so avoid sharing personal information with chatbots. He sees this as a strong case for open source and small models that run locally on your laptop or phone.
Why it matters
OpenAI holds millions of people's conversations. Weak internal controls raise the risk that sensitive data or valuable technology leaks.
What to do
Keep personal and sensitive details out of chatbot conversations, and consider local models for anything private.
Not confirmed yet
All of the Reuters sources are anonymous, and the arrested engineer has not been convicted.

Watch from 6:20 on the episode page

The Loop

The weekly newsletter

The week in AI in one email: what happened, why it matters, what to do about it.