Originally published on The Daily Flare .

OpenAI has restarted training on its frontier AI models after completing a security overhaul, while also reporting a vulnerability in Brotli, a compression technology used across software.

Training restarts after security changes

The return to training marks a new phase for OpenAI's development of its most capable AI systems. The company had paused this work while it carried out a security overhaul. Its resumption indicates that the company considers the changes sufficient to restart the work, though the announcement does not, by itself, explain every measure introduced or how each one will be evaluated over time.

Security is a central concern in frontier model development because training involves valuable research, model weights and computing infrastructure.…