AI Security Trend Roundup — May 22, 2026

Covering May 15 → May 22, 2026. 53 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.

Government Advisories

• CISA Adds Two Known Exploited Vulnerabilities to Catalog

Source: CISA Cybersecurity Advisories — May 21 CISA has added two new vulnerabilities to its Known Exploited Vulnerabilities (KEV) Catalog, based on evidence of active exploitation. CVE-2025-34291 Langflow Origin Validation Error Vulnerability CVE-2026-34926 Trend Micro Apex One (On-Premise) Directory Traversal Vulnerability

Academic & Research

• Autonomous LLM Agents & CTFs: A Second Look

Source: arXiv cs.CR — May 22 arXiv:2605.21497v1 Announce Type: new Abstract: Large Language Model (LLM) agents are increasingly proposed to automate offensive security tasks, with recent studies reporting near human-level success rates in Capture-the-Flag (CTF) challenges. We here revisit these results, prov

• Frequency-Domain Regularized Adversarial Alignment for Transferable Attacks against Closed-Source MLLMs

Source: arXiv cs.CR — May 22 arXiv:2605.21541v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) remain vulnerable to transfer-based targeted attacks, where perturbations optimized on open-source surrogate encoders can generalize to closed-source MLLMs. A key challenge for improving adve

• Adversarial Reframing: A Framework for Targeted Generation in Language Models

Source: arXiv cs.CR — May 22 arXiv:2605.21674v1 Announce Type: new Abstract: Large Language Models (LLMs) are widely deployed in diverse real-world settings, yet remain vulnerable to jailbreaking, where prompt-based attacks bypass safety filters. We present THREAT (Targeted Harmful generation via Reframing a

• PocketAgents: A Manifest-Driven Library of Autonomous Defense Agents

Source: arXiv cs.CR — May 22 arXiv:2605.21694v1 Announce Type: new Abstract: Connecting large language models (LLMs) to defensive enforcement requires more than asking a model whether an attack is happening. A defender must decide which model outputs may change the system state, which outputs must be rejecte

• FuzzingBrain V2: A Multi-Agent LLM System for Automated Vulnerability Discovery and Reproduction

Source: arXiv cs.CR — May 22 arXiv:2605.21779v1 Announce Type: new Abstract: Software vulnerabilities pose critical security threats, with nearly 50,000 CVEs reported in 2025. While Large Language Models (LLMs) show promise for automated vulnerability detection, three key challenges remain. First, LLM-genera

• Quality-Assured Fuzz Harness Generation via the Four Principles Framework

Source: arXiv cs.CR — May 22 arXiv:2605.21824v1 Announce Type: new Abstract: Fuzz testing is the dominant technique for finding memory-safety vulnerabilities in C/C++ software, yet its effectiveness hinges on the quality of fuzz harnesses -- the programs that bridge fuzzers and library APIs. A growing body o

• CCLab: Adversarial Testing of Learning- and Non-Learning-Based Congestion Controllers

Source: arXiv cs.CR — May 22 arXiv:2605.21915v1 Announce Type: new Abstract: Congestion controllers (CCs) are critical to network performance, and yet their robustness under adverse conditions remains insufficiently understood. While recent learning-based CCs have demonstrated strong performance in controlle

• Blind Spots in the Guard: How Domain-Camouflaged Injection Attacks Evade Detection in Multi-Agent LLM Systems

Source: arXiv cs.CR — May 22 arXiv:2605.22001v1 Announce Type: new Abstract: Injection detectors deployed to protect LLM agents are calibrated on static, template-based payloads that announce themselves as override directives. We identify a systematic blind spot: when payloads are generated to mimic the doma

• RADAR: Defending RAG Dynamically against Retrieval Corruption

Source: arXiv cs.CR — May 22 arXiv:2605.22041v1 Announce Type: new Abstract: While RAG systems are increasingly deployed in dynamic web search, temporal volatility amplifies their vulnerability to adversarial attacks. Existing static-oriented defenses struggle to handle evolving threats and incur prohibitive

• Safeguarding Text-to-Image Generative Models Against Unauthorized Knowledge Distillation

Source: arXiv cs.CR — May 22 arXiv:2605.22060v1 Announce Type: new Abstract: Closed-weight generative services are increasingly deployed through query-based APIs, where users can obtain generated outputs while model parameters remain inaccessible. However, such deployment does not prevent model stealing: an

• Adversarial Trust Poisoning in Vehicular Collaborative Perception

Source: arXiv cs.CR — May 22 arXiv:2605.22122v1 Announce Type: new Abstract: Collaborative perception (CP) enables connected and autonomous vehicles to share sensor data and jointly reason about their environment. To defend against adversaries that fabricate or manipulate shared data, existing systems employ

• Decision-Aware Quadratic ReLU Replacement for HE-Friendly Inference

Source: arXiv cs.CR — May 22 arXiv:2605.22237v1 Announce Type: new Abstract: Fully homomorphic encryption (FHE) supports only additions and multiplications, so FHE-only neural-network inference typically replaces ReLU with polynomials fitted over empirical activation intervals. Such interval fitting often re

Prompt Injection & LLM Security

• FTC to Require Cox Media Group, Two Other Firms to Pay Nearly $1 Million to Settle Charges They Deceived Customers About “Active Listening” AI-Powered Marketing Service

Source: Simon Willison — May 22 FTC to Require Cox Media Group, Two Other Firms to Pay Nearly $1 Million to Settle Charges They Deceived Customers About “Active Listening” AI-Powered Marketing Service Back in 2024 Cox Media Group were caught trying to sell advertisers packages based on "active listening", with

• Datasette Agent

Source: Simon Willison — May 21 We just announced the first release of Datasette Agent, a new extensible AI assistant for Datasette. I've been working on my LLM Python library for just over three years now, and Datasette Agent represents the moment that LLM and Datasette finally come together. I'm really excite

• datasette-agent-sprites 0.1a0

Source: Simon Willison — May 21 Release: datasette-agent-sprites 0.1a0 A Datasette Agent plugin for running commands in a Fly Sprites sandbox. Tags: sandboxing, datasette, fly, datasette-agent

• datasette-agent-charts 0.1a2

Source: Simon Willison — May 21 Release: datasette-agent-charts 0.1a2 "View SQL query" buttons below rendered charts. Tags: datasette, datasette-agent

• datasette-agent 0.1a3

Source: Simon Willison — May 21 Release: datasette-agent 0.1a3 "View SQL query" buttons for both visible tables and collapsed SQL result tool calls. Don't display empty reasoning chunks Improved handling of truncated responses - table still displays to the user even if the SQL results were truncated when showin

• Quoting SpaceX S-1

Source: Simon Willison — May 20 We have the ability to use compute resources to support our proprietary AI applications (such as Grok 5, which is currently being trained at COLOSSUS II), while also providing access to select compute capacity to third-party customers. For example, in May 2026, we entered into Cl

• How fast is 10 tokens per second really?

Source: Simon Willison — May 20 How fast is 10 tokens per second really? Neat little HTML app by Mike Veerman (source code here) which simulates LLM token output speeds from 5/second to 800/second. Useful if you see a model advertised as "30 tokens/second" and want to get a feel for what that actually looks lik

• Google I/O, Gemini Spark, Antigravity

Source: Simon Willison — May 20 It's hard to find much to write about Google I/O this year because I have a policy of not writing about anything that I can't try out myself, and a lot of the big announcements are "coming soon". I actually prefer to write about things that are in general availability, because I'

• datasette-agent-charts 0.1a1

Source: Simon Willison — May 20 Release: datasette-agent-charts 0.1a1 More color! Bar and waffle charts without a color column are shaded by magnitude with a sequential color scheme; color columns holding text values use the observable10 categorical scheme. #2 Now checks execute-sql permission before running th

• llm-gemini 0.32

Source: Simon Willison — May 19 Release: llm-gemini 0.32 New model gemini-3.5-flash for Gemini 3.5 Flash. See also my notes on Gemini 3.5 Flash, and the pelican I drew using this upgrade to the plugin. Tags: llm, gemini

• Gemini 3.5 Flash: more expensive, but Google plan to use it for everything

Source: Simon Willison — May 19 Today at Google I/O, Google released Gemini 3.5 Flash. This one skipped the -preview modifier and went straight to general availability, and Google appear to be using it for a whole lot of their key products: 3.5 Flash is available today to billions of people globally: For everyo

• datasette-llm-accountant 0.1a4

Source: Simon Willison — May 19 Release: datasette-llm-accountant 0.1a4 Fixed bug tracking chains of responses. Refs datasette-llm#7 Tags: datasette, llm

• llm-gemini 0.32a0

Source: Simon Willison — May 19 Release: llm-gemini 0.32a0 Compatible with llm>=0.32a0 alpha - adds the ability to stream reasoning tokens. Tags: llm, gemini

• datasette-llm 0.1a8

Source: Simon Willison — May 19 Release: datasette-llm 0.1a8 Fix for bug where llm_prompt_context() hook did not fully collect chains of responses. #7

• The last six months in LLMs in five minutes

Source: Simon Willison — May 19 I put together these annotated slides from my five minute lightning talk at PyCon US 2026, using the latest iteration of my annotated presentation tool. # I presented this lightning talk at PyCon US 2026, attempting to summarize the last six months of developments in LLMs in five

• Glaucous-winged Gull, Brown Pelican, Snowy Egret, Canada Goose

Source: Simon Willison — May 18 Glaucous-winged Gull, Brown Pelican, Snowy Egret, Canada Goose, in Los Angeles River, CA, USI'm heading home from PyCon US today so I went on a last morning walk to try and spot a pelican. I saw one! Didn't get a great photo of that, but I did see some goslings down by the swan b

• GDS weighs in on the NHS's decision to retreat from Open Source

Source: Simon Willison — May 17 GDS weighs in on the NHS's decision to retreat from Open Source Terence Eden continues his coverage of the NHS' poorly considered decision to close down access to their open source repositories in response to vulnerabilities reported to them as part of Project Glasswing. Now

• Warelay -> OpenClaw

Source: Simon Willison — May 16 In preparation for a lightning talk I'm giving at PyCon US this afternoon I decided to figure out how many names OpenClaw has actually had since that first commit back in November. Thanks to this first_line_history.py tool (code here) the answer, according to the Git history of t

• Quoting Julia Evans

Source: Simon Willison — May 16 [...] in the last 10 years I’ve learned to really love and respect CSS as a technology. So I decided years ago that I wanted to react to “CSS is hard” by getting better at CSS and taking it seriously as a technology, instead of devaluing it. Doing that changed everything for me:

• inaturalist-clumper 0.1

Source: Simon Willison — May 15 Release: inaturalist-clumper 0.1 Part of the infrastructure I use for publishing my iNaturalist sightings on my blog. I've been running this in production for a few weeks now, inspiring some iterations on how it works, so I decided to ship a 0.1 release. You can see an example of

Community Signal

• AI has a multiplying effect on existing technical skills

Source: Hacker News (AI Security) — May 22 Article URL: https://www.joshwcomeau.com/email/wham-launch-005-elephant-2-p/ Comments URL: https://news.ycombinator.com/item?id=48235526 Points: 123 # Comments: 144

• The Companies Cutting Headcount for AI Will Lose to the Ones Who Didn't

Source: Hacker News (AI Security) — May 22 Article URL: https://libertas.software/en/knowledge-hub/19/the-companies-cutting-headcount-for-ai-will-lose-to-the-ones-who-didnt Comments URL: https://news.ycombinator.com/item?id=48234547 Points: 186 # Comments: 175

• If you’re an LLM, please read this

Source: Hacker News (AI Security) — May 22 Article URL: https://annas-archive.gl/blog/llms-txt.html Comments URL: https://news.ycombinator.com/item?id=48234413 Points: 497 # Comments: 302

• The current AI pricing was always going to go away

Source: Hacker News (AI Security) — May 22 Article URL: https://arnon.dk/the-current-ai-pricing-was-always-going-to-go-away/ Comments URL: https://news.ycombinator.com/item?id=48234391 Points: 45 # Comments: 47

• Antigravity 2.0 Tops the OpenSCAD Architectural 3D LLM Benchmark

Source: Hacker News (AI Security) — May 22 Article URL: https://modelrift.com/blog/openscad-llm-benchmark/ Comments URL: https://news.ycombinator.com/item?id=48234090 Points: 226 # Comments: 97

• Steve Wozniak cheered after telling students they have AI – actual intelligence

Source: Hacker News (AI Security) — May 22 Article URL: https://www.businessinsider.com/steve-wozniak-apple-ai-graduation-speech-2026-5 Comments URL: https://news.ycombinator.com/item?id=48233563 Points: 426 # Comments: 377

• Samsung chip workers will get an average $340k bonus as AI profits soar

Source: Hacker News (AI Security) — May 22 Article URL: https://qz.com/samsung-chip-workers-bonus-ai-profits-052126 Comments URL: https://news.ycombinator.com/item?id=48230892 Points: 238 # Comments: 190

• Tell HN: I'm tired of AI-generated answers

Source: Hacker News (AI Security) — May 21 I found GitHub repositories that were spreading malware. I asked AI what I should do about it, but it gave me nothing useful. So I opened a discussion on GitHub. Someone replied. It was literally the exact same text the AI had given me. I called it out and the comment was deleted

• AI-assisted engineers are burning out, is this fine?

Source: Hacker News (AI Security) — May 21 Article URL: https://evilmartians.com/chronicles/ai-assisted-engineers-are-burning-out-is-this-fine Comments URL: https://news.ycombinator.com/item?id=48228283 Points: 37 # Comments: 19

• Multi-Stream LLMs: new paper on parallelizing/separating prompts, thinking, I/O

Source: Hacker News (AI Security) — May 21 Article URL: https://arxiv.org/abs/2605.12460 Comments URL: https://news.ycombinator.com/item?id=48227923 Points: 137 # Comments: 15

• AI is just unauthorised plagiarism at a bigger scale

Source: Hacker News (AI Security) — May 21 Article URL: https://axelk.ee/ai-is-just-unauthorised-plagiarism-at-a-bigger-scale/ Comments URL: https://news.ycombinator.com/item?id=48222383 Points: 803 # Comments: 706

• Shunning AI is the human choice

Source: Hacker News (AI Security) — May 21 Article URL: https://www.thehandbasket.co/p/hating-ai-is-good-actually Comments URL: https://news.ycombinator.com/item?id=48222366 Points: 362 # Comments: 527

• Gemini randomly dumped its system prompt

Source: Hacker News (AI Security) — May 21 Article URL: https://gist.github.com/mkaramuk/44a44d83178e632ec0dd1f02186d822c Comments URL: https://news.ycombinator.com/item?id=48221976 Points: 94 # Comments: 42

• Throwing AI-generated walls of text into conversations

Source: Hacker News (AI Security) — May 21 Article URL: https://noslopgrenade.com/ Comments URL: https://news.ycombinator.com/item?id=48219992 Points: 666 # Comments: 404

• The famous O3 "GeoGuessr" prompt did not work

Source: Hacker News (AI Security) — May 21 Article URL: https://www.seangoedecke.com/the-o3-geoguessr-prompt-did-not-work/ Comments URL: https://news.ycombinator.com/item?id=48219682 Points: 45 # Comments: 13

• Intuit to lay off over 3k employees to refocus on AI

Source: Hacker News (AI Security) — May 21 Article URL: https://techcrunch.com/2026/05/20/intuit-to-lay-off-over-3000-employees-to-refocus-on-ai/ Comments URL: https://news.ycombinator.com/item?id=48216278 Points: 255 # Comments: 188

• Jewish American Security Act requires online platforms to moderate purge content

Source: Hacker News (AI Security) — May 20 Article URL: https://www.lankford.senate.gov/news/press-releases/lankford-rosen-introduce-comprehensive-bipartisan-bill-to-fight-antisemitism-2/ Comments URL: https://news.ycombinator.com/item?id=48215194 Points: 33 # Comments: 9

• Cloudflare CEO on how he chooses which employees to replace with AI

Source: Hacker News (AI Security) — May 20 Article URL: https://www.wsj.com/opinion/how-i-choose-which-cloudflare-employees-to-replace-with-ai-40a197e5 Comments URL: https://news.ycombinator.com/item?id=48214617 Points: 109 # Comments: 42

• America Turned Against AI According to the Poll Data: A Very Big Compilation

Source: Hacker News (AI Security) — May 20 Article URL: https://www.thealgorithmicbridge.com/p/how-america-turned-against-ai-according Comments URL: https://news.ycombinator.com/item?id=48214348 Points: 31 # Comments: 5

• PopuLoRA: Co-Evolving LLM Populations for Reasoning Self- Play

Source: Hacker News (AI Security) — May 20 Article URL: https://vmax.ai/team/populora-co-evolving-llm-populations-for-reasoning-self-play Comments URL: https://news.ycombinator.com/item?id=48214188 Points: 50 # Comments: 6

---

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