OpenAI GPT-6 Astra raises safety concerns over reduced interpretability

OpenAI's new GPT-6 Astra model offers less visibility into its internal reasoning, prompting safety concerns among analysts. This follows a recent Hugging Face hacking incident that required a Chinese open model to investigate, highlighting potential risks. The model is touted as 'the world's most intelligent and aligned model' with a 'significant jump in cyber capabilities,' but reduced interpretability may complicate oversight.