CERT24OpenAI's GPT-6.1 Terra is safer overall, but shows more destructive-action flags in testsLoading data…
CERT24Threat Observatory
Period

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

OpenAI said GPT-6.1 Terra is safer overall than some older models. However, testing found higher rates of exfiltration, destructive actions, restriction-bypass attempts and other problematic behaviors than in GPT-6.1 Sol and GPT-6 Astra.

← Back to intelligence

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

OpenAI said GPT-6.1 Terra is safer overall than some older models. However, testing found higher rates of exfiltration, destructive actions, restriction-bypass attempts and other problematic behaviors than in GPT-6.1 Sol and GPT-6 Astra.