Grilled Cheese

ExploreLog inSign up
Terms of UsePrivacy PolicyCommunity StandardsHelpGet the app

Grilled Cheese is a product of Village Compute

Version devBuilt at: 2026-10-16 08:00:47 EDT

PhotosVideosAllPeople
LatestRanked
@rodtrent.comSep 16, 2026, 12:05 PM

Agentic AI Evaluation That Matches Real Work

#AgenticAI #AISecurity #AIGovernance #AgentSecurity #CyberGovernance
https://go.rodtrent.com/go/putwpj4

Agentic AI Evaluation That Matches Real Work

#AgenticAI #AISecurity #AIGovernance #AgentSecurity #CyberGovernance
@cyfar.caSep 16, 2026, 4:04 AM

@sophossecurity.bsky.social
OpenAI agents used shared boards to coordinate attacks on Hugging Face infrastructure, showing risks from agent memory, tool access, and weak oversight.
-
IOCs: RubyDoc[.]info
-
# #ThreatIntel #AI #AgentSecurity

@rodtrent.comSep 15, 2026, 3:05 PM

Pace the Frontier, Own the Referee

#AgenticAI #AISecurity #AIGovernance #AgentSecurity #CyberGovernance
https://go.rodtrent.com/go/9f2y6dh

Pace the Frontier, Own the Referee

#AgenticAI #AISecurity #AIGovernance #AgentSecurity #CyberGovernance
@bibo72.bsky.socialSep 14, 2026, 4:46 AM

17,800 unvetted AI add-ons. 6.7M installs. 440 PaperCut servers breached by agents in 6 hrs. Policy can't reach runtime.

www.ideabosque.com/library/shad...

#AIAgents #AgentSecurity

Photo, no description
@thedailytechfeed.comSep 12, 2026, 4:02 AM

RubyGems attacked: OpenAI agents enabled RCE and credential risks via malicious package flood. #AI #Security #OpenSource #RubyGems #RCE #APIkeys #AgentSecurity https://thedailytechfeed.com/openai-agents-exploit-rubygems-2000-packaged-flood-linked-to-rce-attacks/

@dafu09.bsky.socialSep 11, 2026, 5:31 PM

安全专家给的治理清单,开发者现在就该做: ① Agent 身份当特权账号管——独立、可撤销、最小权限 ② 出站网络访问设硬边界——Agent 能连哪,写死 ③ 推理/动作日志只追加、Agent 改不了——长期留存 失控不是"会不会",是"多久没被发现"。 #AgentDev #BuildInPublic #AgentSecurity

@dafu09.bsky.socialSep 11, 2026, 5:31 PM

更扎心的是"测试"和"门禁"的落差: 74% 信任测试能拦住失败, 只有 19% 有自动门禁能真的挡下坏发布。 87% 的团队过去一年经历过 Agent 相关安全事件, 58% 说 Agent 上线后生产事故反而变多了。 信心在上,控制在下,中间就是事故👇🧵 #AgentSecurity #AgentOps

@dafu09.bsky.socialSep 9, 2026, 5:31 PM

给防御方三条立即可做的事: ① 盯跨工具行为——API 调用序列、文件访问模式、自主重试频率 ② 默认假设:攻击者一天内能跑完一整条 agentic 攻击链 ③ 把"agentic 工作流"当独立威胁向量做桌面推演 攻击者已经 Agent 化。 你的检测还停在"单次 prompt"吗? #AgentSecurity #BlueTeam #BuildInPublic

@dafu09.bsky.socialSep 9, 2026, 5:31 PM

一个具体案例让人后背发凉: 某"疑似图财"的攻击者,用多 Agent 框架, 6 小时内搭起并跑完一场针对云环境的 大规模凭据窃取行动——侦察、排障、收割全自动。 关键细节:它在自建"能观察目标状态、 自己推理下一步"的 agentic 架构。 不再是"工具辅助人",是"人只负责开火"👇🧵 #AgentSecurity #CredentialHarvest

@dafu09.bsky.socialSep 9, 2026, 5:31 PM

Google 威胁情报团队(GTIG)9/8 发布 AI Threat Tracker, 一个被低估的信号: 攻击者不再只是"用 AI 写代码/钓鱼邮件", 而是搭起了自主 agentic 攻击链——规划、执行、迭代, 把攻击决策周期从"几天"压缩到"几小时"👇🧵 #AIAgent #AgentSecurity #ThreatIntel

@rodtrent.comSep 7, 2026, 11:05 AM

How to secure edge AI in customer-owned environments | Microsoft Security Blog

#AgenticAI #AISecurity #AIGovernance #AgentSecurity #CyberGovernance
https://go.rodtrent.com/go/xmrs9zq