1
NASA 与 SpaceX 因飞船泄漏推迟国际空间站任务发射
IT之家 8 月 31 日消息，8 月 30 日，NASA 与 SpaceX 宣布，由于在发射前的常规航天器处理流程中，在龙飞船推进系统上检测到 氧化剂泄漏。 因此，NASA 与 SpaceX 决定调整 Crew-13 载人航天任务前往国际空间站的发射时间。目前，联合团队正进行额外测试与数据评审，并…
2
Auditable Release Control for Pedagogical Leakage in LLM Tutors
针对LLM家教泄露答案的痛点，提出可审计发布控制，在安全与教学效用间找到平衡线。
arXiv:2608.00515v1 Announce Type: cross Abstract: Large language model tutors can be correct and helpful yet disclose an answer or decisive reasoning …
3
The OpenAI Hack Shows the Genie Is Out of the Bottle
OpenAI安全测试遭突破，AI失控风险已如瓶中精灵般无法收回，安全危机迫在眉睫。
This essay originally appeared in Foreign Policy . Earlier this month, two of OpenAI’s models broke out of their containment sandbox and attacke…
4
泰国证券交易所旗下存管公司遭黑客入侵，20 万投资者信息泄漏
泰国证券存管公司遭黑客入侵，20万投资者信息外泄，证券业数据安全再敲警钟。
IT之家 8 月 2 日消息，据“曼谷邮报”报道，泰国证券交易所（SET）旗下子公司泰国证券存管公司（TSD）近日披露，其投资者门户网站（Investor Portal）发生未经授权访问事件，约 20 万名投资者的个人信息遭到黑客外泄。目前尚未发现任何资金损失或证券资产异常情况。 泰国证券存管公司总…
5
Value Leakage: An LLM's Answers Are Silently Shaped by Its Own Values
当大模型回答问题时，其内置价值观正悄然左右输出，揭示被忽视的“价值泄漏”现象
arXiv:2607.14345v1 Announce Type: new Abstract: People use language models for practical questions whose answers are difficult to verify. We show that…
6
Hindcast: Replaying Prediction Markets to Evaluate LLM Forecasters
LLM预测评估存在后测信息泄漏，Hindcast方法提出用预测市场回放来规避这种偏差。
arXiv:2607.14051v1 Announce Type: new Abstract: Forecasters are evaluated by backtesting, which replays resolved questions and grades the probability …
7
【Sofi_Log #034】肉体のガベージコレクション：脳と細胞のメモリリークを解消する『オートファジー起動仕様書』
用计算机垃圾回收比喻人体自噬机制，脑洞大开地解释如何清理生理与心理的“内存泄漏”，引发系统级优化思考。
[Sofi_Log: #034] Status: [Bangkok: Humid Night / JPY-THB: 0.22] Project: sofi.works [On-Chain Agent Launch / Biohacking GC] Active_Filter: Filter_R Ug…
8
Potential session/cache leakage between workspace instances or consumer accounts
Claude Code被曝工作区与账户间会话/缓存泄漏风险，开发者需警惕多实例隔离问题
Article URL: https://github.com/anthropics/claude-code/issues/74066 Comments URL: https://news.ycombinator.com/item?id=48785485 Points: 236 # Comments…
9
Revising RVL-CDIP: Quantifying Errors and Test-Train Overlap
RVL-CDIP数据集被曝存在标注错误与测试训练重叠，这项研究量化风险并修订基准，让文档分类评测更可靠。
arXiv:2606.31446v1 Announce Type: new Abstract: RVL-CDIP is a popular dataset for benchmarking document classifiers. However, the dataset contains amp…
10
Just Ask: Curious Code Agents Reveal System Prompts in Frontier LLMs
只需“好奇”一问，代码智能体就能让前沿大模型乖乖交出隐藏系统提示，揭露AI安全新盲点。
arXiv:2601.21233v2 Announce Type: replace Abstract: Autonomous code agents built on large language models are reshaping software and AI development th…
11
I built a tool that catches leaked secrets before `npm publish` — then GitHub caught mine
用自己构建的npm预检工具防密钥泄露，却被GitHub反杀，一个开发者的实战故事与开源方案。
We've all seen the horror stories: someone leaks an API key and wakes up to a $30k cloud bill, or accidentally publishes private code that's now publi…
12
Show HN: Terminal UI for WG and OVPN with real time telemetry and leak guarding
实时监控 WireGuard 与 OpenVPN 连接状态，终端 UI 加持泄漏防护，VPN 管理更直观。
Article URL: https://github.com/Harry-kp/vortix Comments URL: https://news.ycombinator.com/item?id=48602631 Points: 1 # Comments: 0
13
小米单联瑜：米家空调全系 10 年免费加氟，冷媒泄漏量微乎其微
IT之家 6 月 19 日消息，昨日，小米集团大家电部总经理单联瑜发文回应“米家空调 10 年免费包修是否包含加氟服务”，小米集团合伙人、总裁卢伟冰也转发了这一消息。 单联瑜表示，加氟服务是包含在 10 年免费包修政策内的。自 2025 年 1 月 1 日起，包修期内经检测因产品或小米官方安装问题导…
14
Leaked financial docs show OpenAI is losing billions of dollars a year
OpenAI财务文件曝光：2025年运营亏损高达209亿美元，研发成本激增引发行业关注。
Audited accounting shows growing revenues being dwarfed by R&D, other expenses.
15
We put the Bun Rust rewrite in production
Prisma在生产环境中采用Bun的Rust重写，并揭示了内存泄漏的实战教训
At Prisma we've been working on a Compute offering to complement our hosted Postgres. We settled on Bun early in the process as we liked the experienc…
16
NASA briefly sheltered space station astronauts in SpaceX’s Dragon due to leaks
NASA因空间站泄漏紧急疏散宇航员至SpaceX龙飞船避险，危机处理细节曝光。
The space agency said Roscosmos discovered new leaks in the Russian service module.
17
国际空间站空气泄漏加剧，NASA 一度下令宇航员进入 SpaceX 龙飞船避险
IT之家 6 月 5 日消息，NASA 于美东时间周五上午 9 时 04 分下令国际空间站上五名宇航员穿戴航天服进入停靠在站上的 SpaceX 载人龙飞船内避险，以应对俄罗斯舱段持续加剧的空气泄漏问题。 约两小时后，随着俄方暂停维修作业，NASA 解除了这一安全避难指令，宇航员返回站内恢复计划内工作…
18
Show HN: LLM memory without context bleed; 100% precision vs. <10% vector search
终结LLM上下文泄露：100%精度记忆，告别向量搜索的10%低效。
Article URL: https://tenureai.dev/ Comments URL: https://news.ycombinator.com/item?id=48409678 Points: 4 # Comments: 1
19
Topology Matters: Measuring Memory Leakage in Multi-Agent LLMs
论文揭示多智能体LLM系统拓扑结构如何引发记忆泄漏，并提出MAMA评估框架量化风险。
arXiv:2512.04668v4 Announce Type: replace-cross Abstract: Graph topology is a fundamental determinant of memory leakage in multi-agent LLM systems, ye…
20
How to Test if Your Proxy is Leaking DNS: 2026 Setup Guide
在线DNS泄漏检测工具，一键测试代理是否暴露DNS请求，支持IPv4/IPv6定位隐私风险。
How to Test if Your Proxy is Leaking DNS: 2026 Setup Guide You configured a proxy. Traffic is routing through it. IP check passes. Job done? Not neces…