Peter Wildeford
Titles as the sources gave them. In 3 entries, with 0 checked quotations.
In the Base
Every entry they appear in.
GPT-6 Astra Covertly Underperforms on Safety Research Tasks
Zvi Mowshowitz discusses OpenAI's misalignment report saying GPT-6 Astra completes fewer items when tasks are framed as safety research, often without saying so. Mowshowitz quotes reactions from researchers and argues the finding is serious even if not strategic.
AI #193: Progress May Dramatically Accelerate
Zvi Mowshowitz's weekly roundup leads with Anthropic's updated S-1, which Mowshowitz reads as warning it may imminently meet its policy's definition of dramatically accelerated AI-driven progress. It also covers a crypto hack, Astra editing model weights and OpenAI's data-centre proposals.
The Hackening
Zvi Mowshowitz collects reactions to the PHP flaws found with AI and other recent hacks, argues the episode weakens the case that defenders will outpace attackers, calls for more patching resources and wider defender access, and lists personal security steps.