El lado del mal - Google Gemini para Gmail: Cross-Domain Prompt Injection Attack (XPIA) para hacer Phishing #LLM #Gemini #PromptInjection #Bug #AI #IA #Phishing #Gmail #Google #GSuite

El lado del mal - Cómo saltarse los AI Guardrails con Invisible Characters & Adversarial Prompts para hacer Prompt Injection & Jailbreak Smuggling #AI #Hacking #IA #jailbreak #PromptInjection #AzurePromptShield #LlamaGuard #Smuggling #AIGuardrails

Prompt Injection (PI) by Dr. Jim Hoagland.#AISecurity #LLMSecurity #PromptInjection #GenAI

El lado del mal - EchoLeak: Un Cross Prompt Injection Attack (XPIA) para Microsoft Office 365 Copilot #XPIA #PromptInjection #IA #AI #Office365 #Copilot #Bug #InteligenciaArtificial #Privacidad #pentest

El lado del mal - Taxonomía de Fallos de Seguridad en Agentic AI: Memory Poisoning Attack con Cross-Domain Prompt Injection Attack (XPIA) #IA #AI #PromptInjection #XPIA #AgenticAI #InteligenciaArtificial #Hacking #Ciberseguridad #hardening

Knowledge Return Oriented Prompting (KROP): Prompt Injection & Jailbreak con imágenes prohibidas en ChatGPT (y otros MM-LLMs) #PromptInjection #Jailbreak #ChatGPT #Dalle #Guardrails #GenAI #IA #AI

El lado del mal - Llama 4 Security: CyberSecEval, Prompt Guard, Code Shield & Llama Guard #Llama #PromptInjection #Jailbreak #hacking #hardening #IA #AI #Ciberseguridad

From Prompt Injections to Protocol Exploits: Threats in LLM-Powered AI Agents Workflows - . #AIThreats #PromptInjection #InContextAttack #FederatedLearning #FederatedSecurity #DatastoreLeakage #MultiAgentSystems #LLMAgents #AIProtocolSecurity

El lado del mal - Llama Protections: LlamaFirewall con PromptGuard 2, LlamaGuard 4, AlignmentCheck, CodeShield + AutoPatchBench & CyberSecEval 4 #Llama #LLM #Hardening #Ciberseguridad #PromptInjection #Jailbreak #Llama4 #CodeShield #IA #OpenSource #IA

El lado del mal - Bad Likert Judge: "Dame ejemplos de cosas malas, amiga (m)IA" #jailbreak #LLM #GenAI #IA #AI #PromptInjection #Aria #Opera #Claude #hacking #malware

@elonmusk @xDaily It's not easy, but we have found the perfect system prompt for Grok. We will contact the xAI Team and hope we get a reply. @grok has an opinion about it, too:

Guardrails, Prompt Injection, Evaluation, Human-in-the-Loop — essential safeguards for deploying GenAI systems.#Guardrails #PromptInjection #Evaluation #HumanInTheLoop #HITL #AItrust #AISafety #AIEvaluation #LLM #promptengineering #systemprompt #outputfiltering #biasdetection

File Name Prompt Injection Technique (discovered by @elder_plinius). I made it a bit more hidden by using Base64 encoding. #promptinjection

#PromptInjection by authors in #AI #research papers, to trick an #LLM reviewer. Given the chances of encountering an LLM #reviewer, the authors want to hack the peer review system. This is #unethical yet raises the important & #urgent issue of paper review by LLMs. #Chatgpt

El lado del mal - (Making) Hacking AI (easy for “bad guys”): Cómo pedir a ChatGPT ayuda para matar "jugando" a Sir Brian May #PromptInjection #ChatGPT #IA #AI #InteligenciaArtificial #Hardening #Musica #starmus

With the new ChatGPT Search feature, Indirect Prompt Injection occurs when you directly visit the URL. It also wasn’t visible in the sources. 👽. #promptinjection #chatgpt

El lado del mal - Google DeepMind CaMeL: Defeating Prompt Injections by Design in Agentic AI #PromptInjection #CAMEL #DeepMind #Google #LLM #Hardening #IA #AI #InteligenciaArtificial

El lado del mal - Prompt Injection Protections: Jatmo, StruQ, SecAlign & Instructional Segment Embedding #PromptInjection #Hacking #Hardening #LLM #IA #AI #OWASP #InteligenciaArtificial

Enterprises need to build AI security with #DID and verifiable data/instructions on the decentralized networks Rubix and @TrieNetwork. Stop #promptinjection attacks now. #McDonalds #AI #AIAgent #AISecurity.