Report
Treating 'Super Intelligence' models as insider risks
Satya Nadella argues that controls over model access and actions must sit outside the model itself.
TLDR
Satya Nadella calls for treating frontier models like insider risks—not because they are necessarily malicious, but because they can make mistakes or be compromised. He says their behavior is harder to trace than traditional software and proposes independent access controls, tamper-proof records of their actions, and a way for an authorized person to stop a model mid-task.
Combined views
22.7K
1 Source, first seen ago
380 likes45 comments32 saves12 reposts