Translation in progress Some entries still include text in the original language.

Google 安全 · News

AI Safety HOTOct 12 · Monday

arXivY

Google 安全 · News

Latest publications and developments

Latest publications and developments

Search titles, summaries and full text/

Google 安全

Research topics

Prompt injectionAgent safetyDeception and schemingReward hackingChain-of-thought monitoringInterpretabilityAI controlBackdoors and poisoning隐私与数据泄露Supply-chain securityDangerous capabilitiesAI 与网络攻防开源模型安全Red teaming

Companies and institutions

OpenAIAnthropicGoogle DeepMindGoogle ResearchGoogle 安全MetaxAIMicrosoftNVIDIADeepSeek阿里巴巴 · Qwen智谱 · GLM月之暗面 · Kimi字节跳动腾讯蚂蚁集团上海人工智能实验室UK AISIUS CAISIMETRApollo ResearchRedwood ResearchFAR.AIPalisade Research

Reports and frameworks

System CardFrontier safety frameworks

TimelineTrendingTodaySelected

Google 安全 · News

Latest publications and developments

TimelineTrendingTodaySelected

Google 安全

Research topics

Prompt injectionAgent safetyDeception and schemingReward hackingChain-of-thought monitoringInterpretabilityAI controlBackdoors and poisoning隐私与数据泄露Supply-chain securityDangerous capabilitiesAI 与网络攻防开源模型安全Red teaming

Companies and institutions

OpenAIAnthropicGoogle DeepMindGoogle ResearchGoogle 安全MetaxAIMicrosoftNVIDIADeepSeek阿里巴巴 · Qwen智谱 · GLM月之暗面 · Kimi字节跳动腾讯蚂蚁集团上海人工智能实验室UK AISIUS CAISIMETRApollo ResearchRedwood ResearchFAR.AIPalisade Research

Reports and frameworks

System CardFrontier safety frameworks

Apr 24Fri

Apr 24Friday · 1 items

1. 05:38
   Google Security · Apr 24 05:38Attention5252

   Google 扫描公开网页：间接提示注入的真实分布与增长趋势

   Google 威胁情报团队扫描 Common Crawl 的公开网页，发现真实存在的间接提示注入（IPI）尝试，多数属于无害恶作剧、帮助性引导、SEO 操纵和劝阻 AI 抓取，少数涉及数据窃取和破坏。

Dec 11, 2025Thu

Dec 11, 2025Thursday · 1 items

1. Dec 11, 2025
   Google Bug Hunters · Dec 11, 2025Attention2929

   Task Injection：利用自主 AI 智能体的能动性

   Google 披露了一种针对自主 AI 智能体的 Task Injection 攻击，它与 Prompt Injection 不同，专门利用智能体被赋予的广泛行动与任务执行能力。该攻击对设计用于执行多种任务的 AI 智能体尤为相关。

You have reached the end