AI Security Trend Roundup — Jun 26, 2026

Covering Jun 19 → Jun 26, 2026. 45 new items from 8 tracked sources.

This digest credits every source by name and links directly to each original post. Editorial curation by FixTheVuln — all rights and attribution belong to the original authors.

Academic & Research

• MIRAGE: Protecting against Malicious Image Editing via False Moderation

Source: arXiv cs.CR — Jun 26 arXiv:2606.26199v1 Announce Type: new Abstract: The proliferation of AI-powered image editing systems raises serious concerns because it allows personal images to be arbitrarily manipulated at scale, with minimal effort, and a lower barrier to entry. Prior work on image immunizat

• Data Facts: A Metadata Schema for Structured Data Exchange in the NANDini Multi-Agent Ecosystem

Source: arXiv cs.CR — Jun 26 arXiv:2606.26211v1 Announce Type: new Abstract: NANDini (Networked Agents Natural Distillation of Interconnected Nodal Intelligence) envisions an automated ecosystem where intelligent agents independently create, process, and exchange data to drive decisions at scale. Realizing t

• CyberChainBench: Can AI Agents Secure Smart Contracts Against Real-World On-Chain Vulnerabilities?

Source: arXiv cs.CR — Jun 26 arXiv:2606.26216v1 Announce Type: new Abstract: We present CyberChainBench, a benchmark for evaluating LLM-based agents on smart contract security across three complementary tasks: vulnerability detection, exploit generation, and patch synthesis. Built from 541 real-world exploit

• TEMPO-Diffusion: Temporally Exposed Malicious Poisoning of Diffusion Models

Source: arXiv cs.CR — Jun 26 arXiv:2606.26285v1 Announce Type: new Abstract: Noise-based backdoor attacks on diffusion models typically rely on input-time trigger injection, untargeted activation, and out-of-distribution target generation. Such assumptions reduce both the stealthiness and the practical relev

• Verifying Intent and Harm: A Unified Defense Against LLM-Generated Threats

Source: arXiv cs.CR — Jun 26 arXiv:2606.26377v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly deployed in interactive applications, yet they remain vulnerable to adversarial interactions that induce harmful, deceptive, or policy-violating outputs. Existing defenses typically anal

• Adaptive Evaluation of Out-of-Band Defenses Against Prompt Injection in LLM Agents

Source: arXiv cs.CR — Jun 26 arXiv:2606.26479v1 Announce Type: new Abstract: Recent work (2024 to 2026) has converged on a strategy for defending tool-using LLM agents against indirect prompt injection: rather than training the model to refuse malicious instructions, enforce security outside the model with a

• VIGIL: Runtime Enforcement of Behavioral Specifications in AI Agent Skills

Source: arXiv cs.CR — Jun 26 arXiv:2606.26524v1 Announce Type: new Abstract: Agentic systems increasingly act through third-party skills, allowing model-generated decisions to affect files, communication channels, and cyber-physical devices. These skills often include natural-language specifications that def

• Adversarial Diffusion Across Modalities: A Fusion Survey of Attacks, Defenses, and Evaluation for Text, Vision, and Vision-Language Models

Source: arXiv cs.CR — Jun 26 arXiv:2606.26566v1 Announce Type: new Abstract: Adversarial evaluation of AI systems has matured along four largely disconnected tracks: diffusion-based attacks on text and large language models (LLMs), diffusion-based attacks on image classifiers, jailbreak pipelines against vis

• Agents That Know Too Much: A Data-Centric Survey of Privacy in LLM Agents

Source: arXiv cs.CR — Jun 26 arXiv:2606.26627v1 Announce Type: new Abstract: Large language model agents increasingly query databases, search document collections, call external APIs, remember past interactions, and act on a user's behalf. As they move from answering questions to operating over sensitive dat

• DroidBreaker: Practical and Functional Problem-Space Attacks on Machine-Learning Android Malware Detectors

Source: arXiv cs.CR — Jun 26 arXiv:2606.26707v1 Announce Type: new Abstract: Adversarial APKs are Android applications modified in the problem space to evade machine-learning malware detectors. In this work, we first show that, despite claims, existing problem-space attacks remain largely impractical. Most t

• MIRROR: Novelty-Constrained Memory-Guided MCTS Red-Teaming for Agentic RAG

Source: arXiv cs.CR — Jun 26 arXiv:2606.26793v1 Announce Type: new Abstract: Multimodal agentic retrieval-augmented generation (RAG) systems expand the attack surface beyond prompt injection to include text poisoning, image injection, direct-query attacks, and orchestrator-level tool manipulation. Existing r

• SpikeTimer: Exploring Active Copyright Protection in Spiking Neural Networks via Temporal Backdoor Regularization

Source: arXiv cs.CR — Jun 26 arXiv:2606.26841v1 Announce Type: new Abstract: Spiking Neural Networks (SNN) have emerged as a revolutionary paradigm compared to traditional Deep Neural Networks (DNN) in energy-efficient computing, showcasing exceptional capabilities in processing event-driven sensory data for

• Jailbreaking for the Average Jane: Choosing Optimal Jailbreaks via Bandit Algorithms for Automatically Enhanced Queries

Source: arXiv cs.CR — Jun 26 arXiv:2606.26936v1 Announce Type: new Abstract: With a profusion of jailbreaks for LLMs now widely known, a growing concern is that non-expert malicious actors ("the average Jane") could elicit actionable responses to malicious requests. In this work, we examine whether this conc

Prompt Injection & LLM Security

• AI and Liability

Source: Simon Willison — Jun 25 AI and Liability Bruce Schneier on the recent German ruling that Google be held liable for errors introduced in their AI overviews: AI agents are agents of the person or organization that deploys them—and should be treated by the law as such. If a company hired human writers to w

• datasette-export-database 0.3a2

Source: Simon Willison — Jun 25 Release: datasette-export-database 0.3a2 An embarrassingly tiny release. The pyproject.toml had pinned to datasette==1.0a27, inadvertently making this plugin incompatible with all other Datasette versions. It's now datasette>=1.0a27 instead. Tags: datasette

• simonw/browser-compat-db

Source: Simon Willison — Jun 24 simonw/browser-compat-db Inspired by Mozilla's new MDN MCP service - source code here - I decided to try converting their comprehensive mdn/browser-compat-data repository full of browser compatibility data into a SQLite database. This new GitHub repo includes a Claude Code for we

• Quoting Tom MacWright

Source: Simon Willison — Jun 24 In the last few months, I've started to see [job applications] that were clearly cowritten by an LLM, link to an LLM-generated portfolio site, which then links to LLM-generated GitHub projects, with purely LLM-generated commit messages. [...] My other reaction is that I don't kno

• datasette 1.0a35

Source: Simon Willison — Jun 23 Release: datasette 1.0a35 I'll write more about this one soon, but it's a big release. Three highlights from the release notes: New "Create table" interface in the database actions menu, backed by the //-/create JSON API. It can define columns, primary keys, custom colu

• OPFS + Pyodide test harness

Source: Simon Willison — Jun 23 Tool: OPFS + Pyodide test harness I've been pondering if Datasette Lite - the Python Datasette application run entirely in the browser using Pyodide and WebAssembly - might be able to edit persistent SQLite files stored on the user's computer. That's what OFPS (Origin Private Fil

• Prompt Injection as Role Confusion

Source: Simon Willison — Jun 22 Prompt Injection as Role Confusion First, I absolutely love this: This is a blog-style writeup of the paper. I wish every paper would come with one of these. Academic writing is pretty dry - the impact of a paper can be so much higher if you publish a readable version to accompan

• Porting the Moebius 0.2B image inpainting model to run in the browser with Claude Code

Source: Simon Willison — Jun 22 This morning on Hacker News I saw Moebius: 0.2B Lightweight Image Inpainting Framework with 10B-Level Performance, describing a small but effective inpainting model - a model where you can mark regions of an image to remove and the model imagines what should fill the space. The r

• sqlite-utils 4.0rc1 adds migrations and nested transactions

Source: Simon Willison — Jun 21 sqlite-utils is my combined Python library and CLI tool for working with SQLite databases. It provides an extensive set of higher-level operations on top of Python's default sqlite3 package, including support for complex table transformations, automatic table creation from JSON d

• sqlite-utils 4.0rc1

Source: Simon Willison — Jun 21 Release: sqlite-utils 4.0rc1 See sqlite-utils 4.0rc1 adds migrations and nested transactions. Tags: sqlite-utils

• Temporary Cloudflare Accounts for AI agents

Source: Simon Willison — Jun 21 Temporary Cloudflare Accounts for AI agents The announcement says this is "for AI agents" but (as is pretty common these days) the AI hook isn't really necessary, this is an interesting feature for everyone else as well. Short version: you can now create a Cloudflare Workers proj

• Quoting Sean Lynch

Source: Simon Willison — Jun 19 The real valuable capability MCP offers over skills/CLI is isolating the auth flow outside of the agent’s context window, and potentially out of the harness completely. [...] Maybe the idealized form of MCP is just an auth gateway for the API and nothing else. That’d still be a w

Community Signal

• The AI industry is pouring millions into US elections

Source: Hacker News (AI Security) — Jun 26 Article URL: https://www.bloodinthemachine.com/p/the-ai-industry-is-pouring-hundreds Comments URL: https://news.ycombinator.com/item?id=48687483 Points: 69 # Comments: 35

• The AI backlash is only getting started

Source: Hacker News (AI Security) — Jun 26 Article URL: https://www.economist.com/leaders/2026/06/25/the-ai-backlash-is-only-getting-started Comments URL: https://news.ycombinator.com/item?id=48686219 Points: 68 # Comments: 188

• Why current LLM costs are not sustainable

Source: Hacker News (AI Security) — Jun 26 Article URL: https://aditya.patadia.org/p/ai-and-cloud-costs Comments URL: https://news.ycombinator.com/item?id=48683588 Points: 103 # Comments: 184

• What happened after 2k people tried to hack my AI assistant

Source: Hacker News (AI Security) — Jun 26 Article URL: https://www.fernandoi.cl/posts/hackmyclaw/ Comments URL: https://news.ycombinator.com/item?id=48681687 Points: 291 # Comments: 125

• AI children's books, body horror edition

Source: Hacker News (AI Security) — Jun 26 Article URL: https://lcamtuf.substack.com/p/ai-childrens-books-body-horror-edition Comments URL: https://news.ycombinator.com/item?id=48681250 Points: 199 # Comments: 74

• Hasbro's TV Contracts Ask Child Voice Actors to Sign Rights Away for AI Use

Source: Hacker News (AI Security) — Jun 25 Article URL: https://www.hollywoodreporter.com/business/business-news/studio-minor-performers-surrender-voices-ai-1236630694/ Comments URL: https://news.ycombinator.com/item?id=48678671 Points: 37 # Comments: 7

• The AI Data-Center Boom Is Sparking a Third Wave of Inflation

Source: Hacker News (AI Security) — Jun 25 Article URL: https://www.wsj.com/economy/the-data-center-boom-is-sparking-a-third-wave-of-inflation-926adc6e Comments URL: https://news.ycombinator.com/item?id=48677039 Points: 22 # Comments: 1

• Apple to skip high-end M6 Mac chips in favor of AI-focused M7 line

Source: Hacker News (AI Security) — Jun 25 Giftlink: https://www.bloomberg.com/news/articles/2026-06-25/apple-to-...also https://www.msn.com/en-ca/lifestyle/shopping/apple-to-skip-h... Comments URL: https://news.ycombinator.com/item?id=48676795 Points: 303 # Comments: 329

• Show HN: OpenKnowledge – open source AI-first alternative to Obsidian/Notion

Source: Hacker News (AI Security) — Jun 25 Hi HN, Nick here. We’re launching OpenKnowledge (https://openknowledge.ai/), a “what you see is what you get” markdown editor that has direct integrations with Claude, Codex, and other agents. Available as MacOS app or Web UI+CLI. Fully free/local and OSS.We built this because we

• Ford AI hiccups push carmaker to rehire ‘gray beard’ inspectors

Source: Hacker News (AI Security) — Jun 25 https://archive.ph/DI4Cq Comments URL: https://news.ycombinator.com/item?id=48674446 Points: 594 # Comments: 319

• Zig's new bitCast semantics and LLVM back end improvements

Source: Hacker News (AI Security) — Jun 25 Article URL: https://ziglang.org/devlog/2026/#2026-06-25 Comments URL: https://news.ycombinator.com/item?id=48673825 Points: 266 # Comments: 135

• Political bias in AI: Where the AI models stand

Source: Hacker News (AI Security) — Jun 25 Article URL: https://trakkr.ai/bias Comments URL: https://news.ycombinator.com/item?id=48672779 Points: 160 # Comments: 285

• Why Does Everyone Hate AI?

Source: Hacker News (AI Security) — Jun 25 Article URL: https://paulkrugman.substack.com/p/why-does-everyone-hate-ai Comments URL: https://news.ycombinator.com/item?id=48672694 Points: 86 # Comments: 145

• Ending respiratory infections

Source: Hacker News (AI Security) — Jun 25 Article URL: https://blog.interceptfund.com/p/ending-respiratory-infections Comments URL: https://news.ycombinator.com/item?id=48667588 Points: 200 # Comments: 134

• What I'm Finding About LLM Code Style and Token Costs

Source: Hacker News (AI Security) — Jun 25 Article URL: https://www.jimmont.com/llm-style-token-costs Comments URL: https://news.ycombinator.com/item?id=48667409 Points: 38 # Comments: 17

• Anthropic says Alibaba illicitly extracted Claude AI model capabilities

Source: Hacker News (AI Security) — Jun 24 Article URL: https://www.reuters.com/world/china/anthropic-says-alibaba-illicitly-extracted-claude-ai-model-capabilities-2026-06-24/ Comments URL: https://news.ycombinator.com/item?id=48664814 Points: 777 # Comments: 1268

• Big AI labs are hiring philosophers

Source: Hacker News (AI Security) — Jun 24 https://archive.is/T1FJG Comments URL: https://news.ycombinator.com/item?id=48662452 Points: 152 # Comments: 136

• Exploiting vulnerabilities in Johnson and Johnson web apps

Source: Hacker News (AI Security) — Jun 24 Article URL: https://eaton-works.com/2026/06/24/jnj-webapp-hacks/ Comments URL: https://news.ycombinator.com/item?id=48662347 Points: 95 # Comments: 6

• Ask HN: How much coding should beginners learn in the AI era?

Source: Hacker News (AI Security) — Jun 24 As someone who wants to work in tech in the future, say 5-10 years from now, to what extent do you think coding will be a valuable skill? How much should I learn? Comments URL: https://news.ycombinator.com/item?id=48662310 Points: 33 # Comments: 44

• For most of the world, open-source AI is the only way forward

Source: Hacker News (AI Security) — Jun 24 Article URL: https://techstrong.ai/articles/for-most-of-the-world-open-source-ai-is-the-only-way-forward/ Comments URL: https://news.ycombinator.com/item?id=48660839 Points: 233 # Comments: 144

---

Source List

All sources tracked in this roundup, credited to their original authors/organizations:

• OWASP GenAI Security Project — feed: https://genai.owasp.org/feed/
• Simon Willison — feed: https://simonwillison.net/atom/everything/
• arXiv cs.CR — feed: http://export.arxiv.org/rss/cs.CR
• Protect AI — feed: https://protectai.com/blog/rss.xml
• Google Project Zero — feed: https://googleprojectzero.blogspot.com/feeds/posts/default
• CISA Cybersecurity Advisories — feed: https://www.cisa.gov/cybersecurity-advisories/all.xml
• NIST Cybersecurity News — feed: https://www.nist.gov/news-events/cybersecurity/rss.xml
• Hacker News (AI Security) — feed: https://hnrss.org/newest?q=%22AI+security%22+OR+%22prompt+injection%22+OR+%22LLM+vulnerability%22&points=20