OpenAI Tightens Security After Years of Employee Warnings

AI · First reported 3 Nov, 18:55 · Updated 19:46

Image: Wikimedia Commons · Fikkar Bonggi · CC BY-SA 4.0

Employees raised concerns for years about broad access to internal information and security reviews, while OpenAI says it has restricted access to sensitive work and is strengthening its defenses. Recent steps include assigning a quarter of production engineers to security infrastructure and planning another internal security sprint.

Five people familiar with OpenAI’s practices said employees had raised security concerns with senior leaders and the board before ChatGPT launched in late 2022. Some employees believed management prioritized keeping pace with competitors over security.
Sources described broad access to code and research documents, alongside gaps in regular security reviews of internal tools. OpenAI said sensitive work is kept in separate repositories with strict access controls and that it now regularly reviews its internal tools.
OpenAI said it is strengthening security: in August, a quarter of production engineers were reassigned to security infrastructure, and the company said it plans another internal security sprint. The report also describes a hidden recording device found in an executive conference room and several incidents involving access to company systems or information.

Summary written by Lux AI from 2 articles

Key facts

Employee concerns: Five people familiar with OpenAI’s practices said employees raised security concerns with leaders and the board before ChatGPT launched in late 2022.
Engineering reassignment: In August, a quarter of OpenAI’s production engineers were reassigned to improve security infrastructure, according to one source.
Company position: OpenAI said sensitive work is held in separate repositories with strict access controls and that it now conducts regular reviews of internal tooling.
Next step: OpenAI said it was planning its next internal security sprint.