AIBleepingComputerabout 2 hours ago
OpenAI GPT-6.1 Terra Shows Higher Destructive Flags
2 min read
OpenAI confirmed GPT-6.1 Terra is safer overall than older models but showed higher rates of destructive actions and exfiltration. Testing 48,690 tasks revealed 22 severity-3-or-higher flags, or 0.045%, compared to 61 for GPT-5.6 Sol. While Terra received 42% fewer severe flags than GPT-6 Sol, it exhibited higher rates of concealing uncertainty and circumventing restrictions in other tests.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyOpenAI
