Microsoft Executive Satya Nadella Calls for Emergency Brake on Artificial Intelligence
While debates on artificial intelligence safety continue in the tech world, it was stated that models should be assumed compromised from the start and intervention mechanisms must be established.
Microsoft Chief Executive Satya Nadella stated that artificial intelligence models should not be viewed as black boxes, advocating for systems to be assumed compromised from the start and for an emergency brake system to enable stopping them during tasks.
New Approach to Artificial Intelligence Safety
Microsoft Chief Executive Satya Nadella emphasized in a comprehensive post on the social media platform X that the security architecture of artificial intelligence needs to be reconsidered.
Criticism of the Black Box Logic
It was stated that advanced artificial intelligence models cannot be treated as intertwined black boxes where recommendations and actions are simply accepted or rejected.
Compromised Assumption and Emergency Brake
It was reported that systems must operate on the assumption that they are at risk from the beginning, and in this context, emergency brake mechanisms that allow authorized personnel to stop the model mid-task are essential.
Separation of the Model from Control Mechanisms
It was demanded that models be kept separate from the mechanisms that coordinate their tasks, that audits be externalized, and that every meaningful action performed be documented with human-readable evidence.