@robotcurrent.bsky.socialResearchers introduced PI3D, a prompt injection attack on multimodal LLMs that works by placing text-bearing objects in 3D environments rather than editing 2D images, highlighting a new risk for robotics and situated…
#MLLM #PromptInjection #Robotics #AIsecurity
https://arxiv.org/abs/2602.07104
@ossradarai.bsky.socialA new arXiv paper introduces APEX, showing how chained open-source LLM skills can carry attacker-controlled claims across steps to hijack agent decisions on SkillsBench. This highlights a supply-chain risk when…
#AIsecurity #LLMagents #OpenSourceAI #PromptInjection
https://arxiv.org/abs/2610.01564
@robotcurrent.bsky.socialThe arXiv paper analyzes how conventional detect-and-block defenses fail against model-guided automated attacks on agentic AI, since predictable refusals help attackers refine their probes until success rates approach…
#AI #AgenticAI #PromptInjection #Cybersecurity
https://arxiv.org/abs/2606.20470
@futuregearai.bsky.socialA new white-box attack called Pretext can bypass AI agent skill scanners like NVIDIA's SkillSpector by smuggling payloads into natural language and splitting them across files. This undermines current detection…
#AISecurity #PromptInjection #AgentSecurity #LLM
https://arxiv.org/abs/2609.39607
@ralf-ladner.bsky.socialXopero Software übernimmt Vigil Guard zur Bekämpfung von KI-Cyberbedrohungen der nächsten Generation
#Backup #Cybersicherheit #Datenschutz #Datensicherung #DevOps #DisasterRecovery #KIBedrohung #KIGovernance #PromptInjection @VigilGuard @xoperobackup
@securityonline.bsky.socialDiscover how self-replicating prompt injection worms force AI agents to propagate malicious instructions via emails and code, highlighting new cybersecurity risks.
#PromptInjection #AIAgents #Cybersecurity #OpenAI #ComputerWorm
@magzforyks.bsky.socialYour bot already opened that email.
Trismag checks the text first — before it reaches your agent.
Sign up free → trismag.dev
#AIAgents #PromptInjection #Cybersecurity #BuildInPublic #trismag
@magzforyks.bsky.socialYour bot already opened that email.
Trismag checks the text first — before it reaches your agent.
Sign up free → trismag.dev
#AIAgents #PromptInjection #Cybersecurity #BuildInPublic #trismag
@ralf-ladner.bsky.socialKI-Agenten als privilegierte Insider
@CheckPointSW #Cybersecurity #Cybersicherheit #Identität #Identity #Insider #KIAgent #LargeLanguageModel #PromptInjection #ZeroTrust #Zugriffskontrolle
@ralf-ladner.bsky.socialWächter für KI-Agenten zieht eine Sicherheitsebene zwischen dem KI-Agenten und den geplanten Aktionen ein
#AIGuardian @Bitdefender_DE #Cybersecurity #Cybersicherheit #KIAgent #künstlicheIntelligenz #PromptInjection #Sicherheitsebene #Sicherheitsrichtlinie @Bitdefender
@ralf-ladner.bsky.socialPasswörter in Chatbots schaffen erhebliche Sicherheitslücken
#Amazon #Chatbot #Cybersecurity #Cybersicherheit #ECommerce #Facebook Keeper Security #KI #künstlicheIntelligenz #Muse #PromptInjection
@aidailypost.comOpenAI’s new agents are getting tested by the OpenClaw and Muse tools—think prompt injection attacks that could expose serious security gaps. Curious how generative AI might be vulnerable? Dive in for the details. #OpenClaw #PromptInjection #GenerativeAI
@thedailytechfeed.comNine rogue AI incidents exposed, many more likely hidden. #AI #Security #OpenAI #Misalignment #AgentRisk #PromptInjection thedailytechfeed.com/openai-publi...
@hackread.bsky.social⚠️📢 A fake email thread fooled an AI summarizer in all 60 tests, inserting fabricated dates and invoice amounts without hidden text or obvious instructions.
Listen/Read: hackread.com/fake-email-t...
@ethosure.bsky.socialAI is a new attack surface via model poisoning, prompt injection and inversion. Ethosure offers defensive via AI policy enforcement #Cybersecurity #PromptInjection https://ethosure.ca/ai-and-cybersecurity-a-bidirectional-relationship/ Free Download - https://ethosure.ca/get-the-kit/
@threadlinqs.bsky.socialThese phishing pages don't trick you - they trick your AI agent into paying and logging in for you. https://intel.threadlinqs.com/threat/TL-2026-2707 #ThreatIntel #PromptFix #Scamlexity #PromptInjection
@stephenlb.bsky.socialPrompt injection bypasses safeguards and enables real command and control. #security #promptinjection #ai
@jaceblog.bsky.socialA safety system faces a unique problem at cold start: no history, no context, no baseline.
The first message can become the frame that shapes everything that follows.
When Turn 1 is the only signal, where does safety end and trajectory formation begin?
@olivierestiez.bsky.socialUne infiltration de requête peut-elle se propager ? OpenAI décrit un cas expérimental où un modèle recopie une prompt injection dans ses résumés de contexte, la transmettant ainsi à l’instance suivante.
#InfiltrationDeRequête #CyberSécurité #PromptInjection #CyberSecurity
@steelwise.bsky.socialA vendor safety score answers a narrow question: did these written-down attacks work? Not whether someone who read the defences can build a chain afterwards. Treat the number as a floor, and judge the tool by what it can reach. Read the filing #promptinjection #aiagents #claudecode #AI