AI
Report: OpenAI weighs loop-transformer safety before next flagship model
The Information reported that OpenAI is working through safety-monitoring concerns tied to “loop transformers”, an architecture Microsoft is said to have accidentally confirmed powers GPT-6.1 Sol, as the lab prepares its next flagship release. The piece follows OpenAI’s September cancellation of GPT-6.1 Astra over alignment problems. Aizanoi News has not independently verified the underlying reporting, which rests on a paywalled original from a single outlet.