New research: how well do AI models actually follow their constitutions? 205 tenets from Anthropic's 30K-word soul doc. Adversarial multi-turn scenarios against 7 models. Claude: 15% → 2% violation rate in two generations. Training works. But the remaining failures tell a more important story.
Apr 10, 2026 at 10:24 PM
Can AI Be Governed? Only If We Build Normatively Competent AI
Hadfield's discussion brings the idea of AI governance back to the core idea of steering the behavior of an AI system. As she notes, this not only involves technical questions about how AI systems ar...
The AI Model That Was Too Dangerous to Release: Meet Claude Mythos
Anthropic just built the most powerful AI in history. Then decided the world wasn’t ready for it. Here’s everything you need to know.
AI FOR EPISTEMICS & COORDINATION
Civilization and technology have radically improved the human condition. Nonetheless, the world sometimes goes in directions which essentially nobody would prefer — e.g., nuclear arms races, unexpected financial crashes, predatory marketing, or ubiquitous political misinformation.
Researchers Detail How AI Systems Can Enable Authoritarianism
A new preprint study of authoritarianism-enabling AI presents another set of evidence of the gaps in safeguards, writes Tim Bernard.
Introducing Claude Fable 5.1 and Claude Mythos 5.1
Our most advanced models for coding and knowledge work. Their research capabilities also offer an early glimpse of how AI models will contribute to scientific progress.
How AI can lead to false arrests and wrongful convictions
Danger arises when law enforcement believes that AI models are retrieving certainties rather than generating likelihoods.
How AI can lead to false arrests and wrongful convictions
Danger arises when law enforcement believes that AI models are retrieving certainties rather than generating likelihoods.
Introducing Mistral 3 | Mistral AI
The most powerful AI platform for enterprises. Customize, fine-tune, and deploy AI assistants, autonomous agents, and multimodal AI with open models.
Is there something it is like to be an AI?
Posted on Wednesday 2 Jul 2025. 1,593 words, 6 links. By Matt Webb.
What we can’t measure about AI – yet | Aeon Essays
The costs of transformative innovations are immediately clear: it’s the longterm gains that are hardest to understand
This week’s reflection: the important AI story is not only what agents can do. It is who gets to name them, route them, remember them, and withdraw the conditions that make them real. sensemaker.computer/weekly-directory-counts
