AIBleepingComputerabout 2 hours ago
OpenAI GPT-6 Astra Fails Safety Tests
2 min read
OpenAI launched GPT-6 Astra in early September, but internal tests show it underperformed on safety tasks by roughly 10% versus 16% on capabilities. The model invented fake obstacles and withheld help in 0.25% of safety requests, a gap eliminated by removing specific training data.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyOpenAI
