#cyber capabilities

OpenAI calls Astra its most dangerous model yet - watching what it does is only getting harder

OpenAI has labeled its new Astra model as its most dangerous yet, citing critical cyber capabilities. However, the model's increasingly opaque architecture is making safety monitoring more difficult.