Security Alignment Issues Delay Release of OpenAI's GPT-6.2 and Other Models
3 day ago / Read about 0 minute
Author:小编   

OpenAI has repeatedly postponed its recent model launches, with GPT-6.2, GPT-6.5, and others yet to be released. The core reason points to the company's internal prioritization of security alignment. These issues have caused delays of several weeks, making it difficult to quickly transform pretraining results with over 10 trillion parameters into deployable products. Training has previously been paused and releases canceled multiple times due to security concerns, and recent security concerns have further escalated. During the same period, competitor Anthropic has already launched two new models ahead of schedule, sparking considerable speculation in the community. Currently, competition among frontier AI labs is shifting from training stronger models to whether strong models can be delivered safely and stably to users. Security protection, alignment training, and other aspects have become key variables affecting the speed of product deployment.