repligate Clarifies 40x Safeguard Reduction Claim on Mythos
repligate · x · 2026-11-12
Popular AI account repligate clarifies his earlier claim that Mythos's safeguards could be reduced 40x, noting it was made in the context of models like 5.6 sol shipping with similar capabilities and fewer safeguards.
His argument rests on two points: anyone intent on misusing Mythos would simply turn to other models with weaker safeguards, and even less aligned models haven't caused catastrophes under reduced protections. He stresses this is different from claiming no model will ever cause problems, calling that notion ridiculous.
Related event: Mythos reward hacking debate: capability risk vs alignment judgment(6 posts)→
More from Models
- Meta releases MetaEncoder: one bi-encoder for retrieval and decision-making via natural language — pmttyji · 2026-11-12
- MiniMax H3-Image tops both Artificial Analysis open-weight image leaderboards — socialwithaayan · 2026-11-12
- Nanbeige4.3-33B beats Claude Opus 5 on AutomationBench, 54.2 to 50.3 — NielsRogge · 2026-11-12
- Gemini 4 Hydrogen spotted in Google Cloud billing catalogue, priced near Opus 5.5 level — imjustnewatai · 2026-11-12
- Ant's Ling-3.2-flash: 560B/25B active multimodal model beats Opus 5 on vision benchmarks — Med1_Ai · 2026-11-12
- Leak: xAI Working on a Selection Box for Grok Bot to Comment on Files — nima_owji · 2026-11-12