17h ago

Google DeepMind's Susan Zhang argues AI safety structures let nontechnical actors pursue power without oversight

She claims safety debates fail without global enforcement powers.

0
Original post

"ai safety" has always been an accountability black hole designed for and by the largely nontechnical periphery to become involved in capability development, providing the perfect narrative cover for both sides to run unchecked into the green fields of power they each desire it means absolutely nothing in a hypercapitalistic world where there is no god nor governing body with actual enforcement powers, and any notion of "long-term consequences" is purely a delusional dream for the privileged few who have the luxury of time and resources to theorize, ruminate, and lecture the legal system is too detached, the medical system can't treat, the education system has rotted, and the pretext of democracy and freedom means any centralization of enforcement power either gets false-dichotomied into a false-equivalence with ccp hell or is already doomed at inception to be entirely bought out by those whom they are meant to check-and-balance in the first place

11:46 PM · May 25, 2026 View on X

@suchenzang the industry is capital game and should dissolve for humanity. nothing is worthy.

Susan ZhangSusan Zhang@suchenzang

"ai safety" has always been an accountability black hole designed for and by the largely nontechnical periphery to become involved in capability development, providing the perfect narrative cover for both sides to run unchecked into the green fields of power they each desire it means absolutely nothing in a hypercapitalistic world where there is no god nor governing body with actual enforcement powers, and any notion of "long-term consequences" is purely a delusional dream for the privileged few who have the luxury of time and resources to theorize, ruminate, and lecture the legal system is too detached, the medical system can't treat, the education system has rotted, and the pretext of democracy and freedom means any centralization of enforcement power either gets false-dichotomied into a false-equivalence with ccp hell or is already doomed at inception to be entirely bought out by those whom they are meant to check-and-balance in the first place

6:46 AM · May 26, 2026 · 10.2K Views
10:07 PM · May 26, 2026 · 136 Views

@suchenzang Safety is real in the sense that labs change their behavior because people are like "uh that’s gross" but it’s not their safety people saying it, in fact being a safety person sacrifices your credibility because you're a professional nag so your objection is priced in

Susan ZhangSusan Zhang@suchenzang

"ai safety" has always been an accountability black hole designed for and by the largely nontechnical periphery to become involved in capability development, providing the perfect narrative cover for both sides to run unchecked into the green fields of power they each desire it means absolutely nothing in a hypercapitalistic world where there is no god nor governing body with actual enforcement powers, and any notion of "long-term consequences" is purely a delusional dream for the privileged few who have the luxury of time and resources to theorize, ruminate, and lecture the legal system is too detached, the medical system can't treat, the education system has rotted, and the pretext of democracy and freedom means any centralization of enforcement power either gets false-dichotomied into a false-equivalence with ccp hell or is already doomed at inception to be entirely bought out by those whom they are meant to check-and-balance in the first place

6:46 AM · May 26, 2026 · 10.2K Views
7:34 PM · May 26, 2026 · 416 Views