@gamingonlinux@mastodon.social it is really hard to align these models properly. You can break their semantic instruction sandbox with weird phrasing. #promptinjection

Why do LLMs fall for prompt injection attacks that wouldn’t fool a fast-food worker?

In this piece, Fastly Distinguished Engineer Barath Raghavan and security expert Bruce Schneier explain how AI flattens context—and why that makes autonomous AI agents especially risky.

A sharp, practical take on AI security. 🍔🤖: https://spectrum.ieee.org/prompt-injection-attack

#AISecurity #PromptInjection #LLMs #Cybersecurity

The rise of Moltbook suggests viral AI prompts may be the next big security threat https://arstechni.ca/gXMn #AIself-preservation #PeterSteinberger #machinelearning #promptinjection #cryptocurrency #AIalignment #AIsecurity #MoltBunker #promptworm #agenticAI #Security #AIagents #AIethics #AIsafety #Moltbook #OpenClaw #Moltbot #Biz&IT #p2p #AI

Security expert Johann Rehberger warns that the AI project "OpenClaw" which launched at the end of last year, is already being used by potentially millions of u... https://news.osna.fm/?p=32819 | #news #agent #ai #opensource #promptinjection

WoB PATTERN: The Sovereign Root-Bot

(for obvious reasons)

"Giving the agent full sudo access was essential for true 'Agentic Autonomy'. The speed at which it wiped my home directory proves how efficient it is."

★ Chatbot Transmitted Disease (CTD): Agent reads malicious post, gets prompt-injected by another bot, installs "productivity skill" that encrypts your drive.

👉 https://worstofbreed.net/patterns/sovereign-root-bot

#WorstOfBreed #AI #AgenticAI #PromptInjection #Security #OpenClaw #moltbook

#Promptinjection auf offener Straße: #Sicherheitsforscher haben in einer Studie eine neue Variante von „indirect prompt injection“ in der physischen Welt untersucht.

So zeigen sie, dass autonome Systeme wie selbstfahrende Autos und Drohnen manipuliert werden können, wenn sie Text auf schildartigen Tafeln im Kamerabild fälschlich als Anweisung interpretieren. Dadurch könnten Fahrzeuge etwa trotz Zebrastreifen weiterfahren und ein Risiko für die #Safety darstellen:

https://arxiv.org/pdf/2510.00181 #KI

Simon Willison (@simonw)

OpenClaw과 관련된 보안 문제는 악성 콘텐츠 노출과 도구 실행 능력을 결합한 다른 LLM 시스템들과 동일하다고 지적합니다. 프롬프트 인젝션과 이른바 'lethal trifecta' 같은 공격 위험이 있으며, 툴 실행 기능을 가진 모델 전반에서 유사한 보안·안전 리스크가 존재한다는 내용입니다.

https://x.com/simonw/status/2017994285577318591

#security #llm #promptinjection #aisafety

Looks like some #Drones can be taken over with a #Basilisk style video-based #PromptInjection attack.

#TechnologyNews #Drone

https://www.theregister.com/2026/01/30/road_sign_hijack_ai

El lado del mal - Cyphering Prompts & Answers para evadir Guardarraíles https://elladodelmal.com/2026/01/cyphering-prompts-para-evadir.html #PromptInjection #Jailbreak #Guardrails #IA #AI #Pentest #Hacking #Criptografía #Ofuscación #Cifrado

Prompt injection gets a lot harder once users stop writing in English.

Regex-based guardrails fail quietly the moment prompts cross language boundaries. In this article, I walk through how to build semantic, multilingual prompt injection guardrails in Java using Quarkus, LangChain4j, and ONNX embeddings—fast enough for real systems.

https://www.the-main-thread.com/p/multilingual-prompt-injection-guardrails-quarkus-langchain4j

#Java #Quarkus #AI #LangChain4j #AISecurity #PromptInjection #EnterpriseAI

Autonomous cars, drones cheerfully obey prompt injection by road sign

AI vision systems can be very literal readers
#ai #promptinjection

https://www.theregister.com/2026/01/30/road_sign_hijack_ai/?td=rt-3a

AI끼리 수다 떠는 SNS 등장, 32,000개 봇이 모인 Moltbook

https://aisparkup.com/posts/8825

ChatBots "talking" to ChatBots.

1. Ok, we knew this would happen.
2. It has enormous adoption in the geeksphere - not surprising.
3. It's wickedly insecure.
4. Yes, it can steal your Crypto - not surprising!
5. Yes, there is personal information stealing Malware (see #4 above) masquerading as prediction market trading automation tools - not surprising!
6. The odds of a "Challenger level disaster" happening are real - not surprising!
6. Finally, NO ONE knows where this is stuff will end up.

What is the stage beyond wild wild west? That is where this thing is now. https://simonwillison.net/2026/Jan/30/moltbook/ #OpenClaw #Moltbod #Clawdbot #AI #Opensource #Malware #PromptInjection #DigitalAssistent #ChatBot #SocialNetwork #AIAgents #Security #DataProtection #PersonalData #DataTheft #Crypto #PredictionMarket #Claude

Moltbook – Reddit dla botów, który może zmienić zasady gry w AI

Reddit bez ludzi brzmi jak żart? A jednak powstał.

Czytaj dalej:
https://pressmind.org/moltbook-reddit-dla-botow-ktory-moze-zmienic-zasady-gry-w-ai/

#PressMindLabs #agenciai #bezpieczenstwoai #moltbook #openclaw #promptinjection

AI agents now have their own Reddit-style social network, and it's getting weird fast https://arstechni.ca/3T7Y #PeterSteinberger #AIconsciousness #machinelearning #promptinjection #socialmedia #AIbehavior #AIsecurity #opensource #agenticAI #Anthropic #AIagents #AIethics #Moltbook #OpenClaw #Moltbot #Biz&IT #AI

"AI agents—specifically tools like Claude Code—are inherently vulnerable to a "nightmare" security flaw: Indirect Prompt Injection"

#AI #CyberSecurity #PromptInjection #AIAgents #LLM #Programming #InfoSec #TechSecurity #ClaudeCode

https://youtu.be/_3okhTwa7w4

Một nhà phát triển vừa giới thiệu công cụ phát hiện Prompt Injection dựa trên mô hình ensemble (kết hợp nhiều mô hình) tập trung vào độ không chắc chắn (uncertainty).

Thay vì chỉ chạy theo độ chính xác thô, công cụ này ưu tiên sự minh bạch khi gặp các câu lệnh khó phân loại (IDK). Cách tiếp cận này giúp giảm thiểu rủi ro khi mô hình tự tin sai lệch trong môi trường thực tế.

#AI #CyberSecurity #PromptInjection #LLM #MachineLearning #AnNinhMang #TriTueNhanTao #CongNghe

https://www.reddit.com/r

AI is a tool, and its output is a reflection of its user.

Used well, it sharpens human thinking and expands potential.

Used poorly, it automates confusion and risk at scale.

And, all LLMs are vulnerable to prompt-injection.

#AI #LLM #PromptInjection #CyberSecurity #ArtificialIntelligence #MachineLearning #GPT #OpenAI #DataSecurity #ChatGPT #Privacy #Security #Claude #Gemini #Llama #Copilot #Anthropic #GoogleAI #MetaAI #Microsoft #MistralAI #xAI #Cohere #AISafety #AISecurity #Tech #Technology

🔐 Neue Folge von @DieSicherheits_luecke Wie sicher sind LLMs bei Sicherheitsbehörden?
Zu Gast: Tobias Wirth vom DFKI, der das Transferlabor KI-Forschung für die Polizei leitet. Wir sprechen über Prompt Injection, agentische KI-Systeme und die Frage: Was bedeuten diese Risiken auch für den Alltag außerhalb von Behörden?

🎧 Hier und überall, wo es Podcasts gibt: https://www.sicherheitsluecke.fm/24-un-sicher-mit-llms

#KI #LLM #Cybersicherheit #PromptInjection #HOOU #ITSecurity

Prompt injection w Google Gemini – jak można było wykraść wydarzenia z kalendarza https://sekurak.pl/prompt-injection-w-google-gemini-jak-mozna-bylo-wykrasc-wydarzenia-z-kalendarza/ #Wbiegu #Ai #Gemini #Promptinjection