OpenAI's GPT-6 Astra reduces hallucinations but still vulnerable to hidden prompt injections

openai's new model, gpt-6 astra, hallucinates less than gpt-5.6 sol and blocks 99.99 percent of direct prompt injection attacks. however, it remains vulnerable to indirect prompt injections, with a failure rate of 8.5 percent. astra also resists jailbreaks more effectively, but persistent attackers can still get problematic responses. openai's gpt-red method is credited for the improved defense rate. the model's security risks have implications for enterprise ai agent deployments.