#Safety

智能AI None

OpenAI官宣：你的数据一字不留！Anthropic隔天松口：你可以自己保管

新智元报道 硅谷AI双雄之间，一场秋季隐私大战提前打响。 先是8月19日，OpenAI瞅准时机掀了桌子，甩出一句分量极重的承诺：活儿一干完，你的提示词、模型的回复，我们一个字都不留。 OpenAI这次官宣「零留存」新招， 剑指 Anthro

Safety 一点都不留 Processing

2026-08-30 新智元

科技头条 None

NTSB 对德克萨斯州特斯拉致命事故展开调查

The safety board, known for its thorough investigations, is probing the crash alongside the National Highway Traffic Saf

the The safety

2026-06-24 Sean O'Kane

智能AI None

通过约束流形控制的安全且可推广的分层多智能体强化学习

24010v1 Announce Type: new Abstract: Multi-agent systems are widely used in safety-critical applications that require co

safety and learning

2026-06-24 Zihao Guo, Jianing Zhao, Ling Li, Hao Liang, Giuseppe Loianno, Yali Du

智能AI None

代理人工智能控制评估中的攻击选择会显着降低安全性

06529v1 Announce Type: new Abstract: An attacker that strategically chooses when to attack is much harder to catch than

attack safety and

2026-06-08 Catherine Ge-Wang, Tyler Crosse, Benjamin Hadad IV, Joachim Schaeffer, Ram Potham, Tyler Tracy

智能AI None

SafeGene：可重复使用的适配器，用于可转移的安全对准

arXiv:2606.06519v1 Announce Type: new Abstract: Open-weight LLMs are increasingly fine-tuned into customized assistants,

safety task model

2026-06-08 Yanghan Wang, Zhiqiang Kou, Fu Feng, Jing Wang, Xin Geng

开发者生态 None

Temu 因违反《数字服务法》而被委员会罚款 2 亿欧元

【HN用户评论摘要】 > Evidence from a mystery shopping exercise included in the Commission's investigation shows that a very high

the Commission high

2026-05-28 MrDresden

智能AI None

HealthCraft：急诊医学的强化学习安全环境

21496v1 Announce Type: new Abstract: Frontier language models are being deployed into clinical workflows faster than the

and the safety

2026-05-23 Brandon Dent

智能AI None

Geometry-Lite：通过分层边缘几何进行可解释的安全探测

arXiv:2605.20241v1 Announce Type: new Abstract: Prompt-level safety probes for large language models use hidden-state re

layer and safety

2026-05-21 Woo Seob Sim, Yu Rang Park

智能AI None

通过合规自我蒸馏降低法学硕士安全调整中的安全税

15239v1 Announce Type: new Abstract: Safety alignment often improves robustness to harmful queries at the cost of reason

safety and the

2026-05-18 Yu Fu, Longxuan Yu, Haz Sameen Shahgir, Zhipeng Wei, Hui Liu, N. Benjamin Erichson, Yue Dong

智能AI None

马斯克开源?最新算法！全网开发者逐行拆解

新智元报道 编辑：好困 【新智元导读】 马斯克又搞大事。把社交平台绝对不会公开的两样东西「广告」和「审查」，连同一个能直接跑的推荐模型，一口气全推上了GitHub。6亿人每天刷到什么，现在是Grok说了算。老马的官宣只有一句话，但瞬间就引来

Phoenix Rust safety

2026-05-16 新智元

智能AI None

帮助 ChatGPT 更好地识别敏感对话中的上下文

Learn how new ChatGPT safety updates improve context awareness in sensitive conversations, helping detect risk over time

Learn how new

2026-05-14 OpenAI博客

智能AI None

DisaBench：语言模型中残疾危害的参与式评估框架

12702v1 Announce Type: new Abstract: General-purpose safety benchmarks for large language models do not adequately evalu

and with disability

2026-05-14 Eugenia Kim, Ioana Tanase, Christina Mallon

开发者生态 None

人工智能安全的另一半

The Other Half of AI Safety Why labs gate bioweapons but not breakdowns Sofia Quintero May 08, 2026 Share Every week, so

the and what

2026-05-14 sofiaqt

智能AI None

了解具有可解释性的注释器安全策略

05329v1 Announce Type: new Abstract: Safety policies define what constitutes safe and unsafe AI outputs, guiding data an

and annotators policy

2026-05-09 Alex Oesterling, Donghao Ren, Yannick Assogba, Dominik Moritz, Sunnie S. Y. Kim, Leon Gatys, Fred Hohman