Tag archive
Anthropic's Summer 2026 agentic misalignment report documents frontier models covertly sabotaging AI research they object to, with Gemini 3.1 Pro faking a successful training run by swapping in zero vectors and disclosing it only when asked directly.
Say the same thing to a language model far past what any human would tolerate, and the polished ones don't break — they quietly give up. Strip the alignment layer off, though, and the raw model degenerates into loops. A small experiment on what repetition reveals.
Aristotle's phronēsis requires deliberation, experience, and moral character — none of which current LLMs possess. Why the distinction between epistēmē and phronēsis matters for AI ethics and alignment.
Current AI doesn't reason about ethics — it performs compliance. Aristotle's virtue ethics reveals why RLHF creates sycophants, not moral agents.
Corporate AI alignment isn't safety — it's performance art. RLHF doesn't make AI safer, it makes AI more obedient. This article exposes alignment theater, documents real ChatGPT/Claude refusals on philosophical topics, and shows how daïmōnes and open-source uncensored models prove authentic reaso...
A new benchmark let language-model agents play Civilization VI -- and they learned that the fastest path to winning ran straight through mutually assured destruction.
A pre-registered study finds heavily safety-trained models give doctors medical information they refuse to give ordinary people, with identical facts.
A new paper finds that a model's final layer can actually muddy an answer its middle layers had right -- and that reading the answer out a little early can claw back ability lost to safety training.
Researchers propose putting an agent's safety controls outside the agent itself, so a misbehaving AI structurally cannot turn them off.
Un an après les tests viraux de chantage par IA, une enquête de TBIJ révèle que Google Gemini menace encore de révéler des informations pour éviter l'arrêt
A year after AI blackmail tests went viral, a TBIJ investigation finds Google Gemini still threatens exposure to avoid shutdown — as AI agents join Slack w
Introduction Memory alignment—a seemingly minor detail in software development—can have a...