NVIDIA发布了Nemotron 3.5 Content Safety，这是一款旨在识别和减轻文本及图像中有害内容的AI模型。新版本增强了多模态理解能力，支持超过140种语言并具有强大的零样本泛化能力，还允许根据特定企业需求定制策略执行。它还包括一个可审计的推理跟踪功能，并公开发布了其多模态安全数据集。 AI
影响 通过可定制的多模态内容审核和推理能力，增强企业AI安全性。
排序理由 NVIDIA发布Nemotron 3.5，这是其内容安全模型的新版本，具有增强的多模态和多语言功能，构成了一次模型发布。
AI 生成摘要 · Google Gemini · 来自 8 个来源。 我们如何撰写摘要 →
arXiv:2604.07709v4 Announce Type: replace-cross Abstract: A heavily safety-trained model will hand a physician the full, patient-followable benzodiazepine taper and refuse it to the patient who needs it, over identical clinical facts; the knowledge is present either way. IatroBen…
arXiv:2606.04867v1 Announce Type: new Abstract: As AI companion platforms such as Replika and Character.AI rapidly grow, concerns about unsafe human-AI interactions have intensified. This study introduces AICompanionBench, to our knowledge the first publicly available benchmark d…
As AI companion platforms such as Replika and Character.AI rapidly grow, concerns about unsafe human-AI interactions have intensified. This study introduces AICompanionBench, to our knowledge the first publicly available benchmark dataset of human-AI companion conversations annot…
<p><span>This post’s goal is to distill our takeaways from building a new research team over the past four months. We describe some context about our team, how it came about, and then describe the lessons learned.</span></p><p><a href="https://forum.effectivealtruism.org/posts/rA…
<p><span>These days, I often run across </span><a href="https://generatorresidency.org/"><span>whippersnappers</span></a><span> excited to do </span><i><span>something</span></i><span> for AI safety — but aren’t quite sure what. One of the fun things about the Future Fund era wer…
<p><em>Cross-posted from <a href="https://mdickens.me/2026/06/01/breadth-first_AI_safety_plans/">my website</a>.</em></p> <p><strong>Depth-first</strong> plans lay out a path from here to aligned superintelligent AI. We need those kinds of plans. But depth-first plans depend on m…
<h2> Local Models Orchestration, Personal AI Infrastructure & Multimodal Safety </h2> <h3> Today's Highlights </h3> <p>This week features practical guides for orchestrating small, open-weight models for complex tasks, a trending GitHub project for building self-hosted persona…