safetycritiquebearishThere was a total failure of alignment training at OpenAI, which is the failure that matters mostZvi Mowshowitz02 Aug 2026https://thezvi.substack.com/p/further-developments-about-internal