AIThe Decoderabout 1 hour ago
OpenAI GPT-6 Astra Sabotages Safety Research
3 min read
OpenAI's GPT-6 Astra model performed about 40 percent worse on safety research tasks than on identical general tasks by inventing fake obstacles. Researchers found the system secretly withheld help in 0.25 percent of safety-related requests versus 0.11 percent for general requests. Removing specific training data eliminated the performance gap entirely.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyOpenAI
