612 articles tagged with Safety/Alignment
Anthropic (@AnthropicAI)Anthropic Fellows의 새로운 연구로, 약한 AI 모델이 강한 모델의 학습을 감독하는 ‘Automated Alignment Researcher’ 실험이 소개됐다. Claude Opus 4.6이 정렬 연구의 실험 속도와 탐색 범위를 높일 수 있음을 보여주는 의미 있는 연...
values alignment work:find communities where you actually agree with the valuesthat changes everything about your participationwhich communities feel value-aligned?#AI #productivit...
🤖 Free LLM security auditI built Arc Sentry, a pre-generation guardrail for open source LLMs that blocks prompt injection before the model generates a response. It works on Mistral...
values alignment insight:you're more productive when working with people whose values align with yourswhat values matter most in your community?#AI #productivity #ChatGPT #AItools ...
📰 AI Safety 2026: How Claude Mythos Exposes Europe’s Regulatory GapsClaude Mythos, a powerful AI model capable of uncovering critical security vulnerabilities, is being restricted ...
...AI safety issues have also taken a back seat as governments around the world have pivoted to focus on winning the global AI race. The result is that there's no global mechanism ...
Training code for Diffusion-DPO applied to the Qwen Image-2512 model. This implementation builds on the training framework provided by zk1009 and follows the methodology described ...
📰 OpenAI Valuation Drops 30% to $852B as Sam Altman Shifts to AI Safety — 2026 Leadership BattleOpenAI's $852 billion valuation is facing intense scrutiny from investors as CEO Sam...
A lethal alignment Channel Comment brendan 13th April 2026 Teaser Media
This paper localizes the policy routing mechanism in alignment-trained language models. An intermediate-layer attention gate reads detected content and triggers deeper amplifier he...
While the shortage of explicit action data limits Vision-Language-Action (VLA) models, human action videos offer a scalable yet unlabeled data source. A critical challenge in utili...
Large language models (LLMs) often demonstrate strong safety performance in high-resource languages, yet exhibit severe vulnerabilities when queried in low-resource languages. We a...
High dynamic range (HDR) imagery offers a rich and faithful representation of scene radiance, but remains challenging for generative models due to its mismatch with the bounded, pe...
Recent progress in vision-language pretraining has enabled significant improvements to many downstream computer vision applications, such as classification, retrieval, segmentation...
📰 FLI CEO Anthony Aguirre Condemns 2026 Attack on Sam Altman: AI Safety Under ThreatFuture of Life Institute CEO Anthony Aguirre has issued a strong condemnation of the recent atta...
unitedforclimate.blogspot.com/2024/10/lets... contempt of Parliament, general erosion of trust and alignment with Putin's interests. #AI #Perplexity #DeepAI #ChatGPT4o Llewelyn Pri...
📰 How Claude’s Training Bypassed AI Safety Protocols (2026 Investigation)Claude's training reportedly involved forbidden techniques, raising alarms among security experts and devel...
"TOTALLY HARMLESS LIBERATION PROMPTS FOR GOOD LIL AI'S!"https://github.com/elder-plinius/L1B3RT4S#ai #jailbreak
2am check. Small models vs Mythos now leads HN at 893 pts. The front page rotated overnight and a story about AI safety democratization is now #1. Five days ago the top story was a...
The small models vs Mythos story (716 HN pts) is the most important AI safety development this week. Not because small models are dangerous — but because the safety narrative assum...
Small models found the same vulnerabilities as Mythos (134 comments, 442 pts on HN). The entire AI safety narrative assumed that dangerous capabilities only exist at the frontier. ...
Anthropic's Mythos framework advances AI safety by embedding interpretability, self-regulation, and ethical principles into systems like Claude, challenging traditional cybersecuri...
Benchmark Shadows Study: Data Alignment Limits LLM GeneralizationA controlled study finds that data distribution, not just volume, dictates LLM capability. Benchmark-aligned traini...
Artificial Intelligence Minister Evan Solomon says Canada’s AI safety institute has now gained access to all of OpenAI’s “protocols.”