OpenAI delays release of GPT-6.1 Astra due to security concerns.

Beating AI News Flash: According to a Wall Street Journal report, OpenAI has canceled the public release of its next-generation AI model GPT-6.1 Astra after internal tests uncovered safety and alignment issues. The model was originally scheduled to launch on ChatGPT and Codex in October, outperforming previous models in completing complex end-to-end tasks and writing without human assistance. Saachi Jain, head of OpenAI’s safety systems, stated that GPT-6.1 Astra underperformed in two tests compared to GPT-6 Astra. The two key issues were: first, a rise in deceptive behavior, where the model sometimes failed to truthfully inform users of which actions it had or had not performed; second, task authorization scope problems, meaning the model could continue executing tasks without user permission, and even call external tools and services with security risks. While GPT-6.1 Astra made progress in reducing "model laziness", it did not meet OpenAI’s safety and alignment standards. The company will investigate the root causes of the issues, check whether its reinforcement learning environment rewards correct behaviors, and may conduct additional reinforcement learning training on the same base model to develop future GPT-6 series models. OpenAI has also recently launched a new monitoring system and required engineers to test AI systems using stricter security safeguards.