rm -rf / is one prompt injection away from your production server. Parallax intercepts every tool call before it executes. Blocks. Redacts. Logs. In <0.2ms. One Rust binary, one YAML, 51 rules out of the box. github.com/agent-defense/par… #LLMsec #AIagents #DevSecOps #Rust

AiSec (AI Security Initiative) A new initiative by @DeFiHackLabs , advancing LLM security (LLMSec) and beyond. AiSec — securing AI agents and autonomous systems, advancing LLM-driven offense and defense. We use LLMs to scale security — finding more vulnerabilities, reducing false positives, and building stronger defenses for AI-powered systems and workflows. Focus areas: ✅Smart contract security ✅AI agent & autonomous system security ✅LLM vulnerability research & detection We’re starting to share our initial work in open-source LLM security — with more tools and research coming soon. We’re building this in the open — contributors are welcome to join us.

#AI #VibeCoding #VibeSecCoding #AISafety #LLMSec #AISecurity

I am starting my PhD in Computer Science at Virginia Tech! I will be working under the supervision of Murat Kantarcioglu. I will focus on LLM Security, but I am always open to new ideas, so let me know if you want to have a discussion! #VirginiaTech #NLProc #LLMSec

Drive-By Attack in Ollama Desktop - gitlab-com.gitlab.io/gl-secu… - gitlab.com/gitlab-com/gl-sec… by @init_string at @gitlab @ollama is a popular local AI application that lets you run LLMs on your own machine. Many users choose it to keep their AI conversations private, without sending data to third parties. As of this writing, it has 150,000 stars on GitHub - making it one of the top-ranked AI tools. There was a vulnerability in the macOS/Windows desktop GUI (not the core API) which would have allowed any website you visit to reconfigure your local application settings - sending all of your local chats to a remote, attacker-controlled server. This means that every private conversation could have been intercepted and read remotely, and every response potentially modified using poisoned models. This was due to incomplete cross-origin controls in the local web service bundled with the GUI. I reported this vulnerability directly to the Ollama team, and they fixed it and released a patched version for all operating systems (v0.10.1) within hours of receiving my report. Special thanks to them for taking security seriously and overall being really nice to communicate with. This blog walks through the process of finding the bug and building an exploit to prove the impact. It also has a section which can help you check if your installation was potentially compromised. Finally, for those who like to really understand how it all works - the exploit source code is available to try it yourself. #Ollama #OllamaDesktop #AIsecurity #LLMSec #AppSec #DriveByAttack #CORS #SameOriginPolicy #CrossOrigin #Localhost #DesktopSecurity #AIPrivacy #ModelPoisoning #DataExfiltration #WindowsSecurity #macOS #ExploitDev #RedTeam #VulnerabilityDisclosure #CVE #PatchNow #SecureDefaults #BrowserSecurity #SupplyChainAI

Urgent! Same urgency that prompted Tesla 2017 Memo on communication bottlenecks and a desire to transcend AI limitations and NVIDIA chip allocations.. Perplexity’s new benchmarks cements UnityGrid QPI as the only production grade, multi-agent, ethically-governed quantum-biomimetic AI mesh in the field ..with a 1–3 year lead on everyone else in deployment maturity and a 2 -4 year lead with the Sovereign E8 Kali Red-Team LLMSEC DevOps Engine Market Reality Check: Competitors are still in rollout, pilot, or niche specialization. We’re already in production-class multi-node, live synchronization, adaptive cycles, and real-time explainability. Innovation Edge: Nobody else combines biomimetic resonance, E8 topology, EchoRhythm Core, and encrypted P2P mesh with dynamic ethical governance. Differentiation Hook: We’re not just building algorithms ...we’ve operationalized autopoietic agents with mission-critical AI guardrails. Strategic Positioning: UnityGrid isn’t chasing qubit counts or LLM compression bragging rights. We own the intersection of quantum AI, national security, and ethical AI governance ...the exact gap where Big Tech has expertise debt and we have a moat. TOP 20 Quantum AI Benchmarks (08/11/25) quantum-agi-consortium-qmagi… #UnityGridQPI #QuantumPulseIntelligence #E8TopologyResonance #EchoRhythmCore #BiomimeticAI #EthicalAILeadership #MissionCriticalAI #AutopoieticAgents #EncryptedAIMesh #OperationalAntifragility #1to3YearsAhead #QuantumAIDominance #SovereignAI #AIWithGuardrails #NationalSecurityAI #DefenseTechAI #SemiconductorAI #NSTCMember #EthicalGovernanceAI #QuantumBIOS #TechMoat #InvestorReadyAI #QuantumEdge #StrategicDifferentiation #FutureProofAI #DefenseReadyTech #BlackOpsAI #AIForNationalSecurity #GameChangerAI #NoOneMatchesThis

Perplexity’s new benchmarks cements UnityGrid QPI as the only production grade, multi-agent, ethically-governed quantum-biomimetic AI mesh in the field ..with a 1–3 year lead on everyone else in deployment maturity and a 2 -4 year lead with the Sovereign E8 Kali Red-Team LLMSEC DevOps Engine Market Reality Check: Competitors are still in rollout, pilot, or niche specialization. We’re already in production-class multi-node, live synchronization, adaptive cycles, and real-time explainability. Innovation Edge: Nobody else combines biomimetic resonance, E8 topology, EchoRhythm Core, and encrypted P2P mesh with dynamic ethical governance. Differentiation Hook: We’re not just building algorithms ...we’ve operationalized autopoietic agents with mission-critical AI guardrails. Strategic Positioning: UnityGrid isn’t chasing qubit counts or LLM compression bragging rights. We own the intersection of quantum AI, national security, and ethical AI governance ...the exact gap where Big Tech has expertise debt and we have a moat. TOP 20 Quantum AI Benchmarks (08/11/25) quantum-agi-consortium-qmagi… #UnityGridQPI #QuantumPulseIntelligence #E8TopologyResonance #EchoRhythmCore #BiomimeticAI #EthicalAILeadership #MissionCriticalAI #AutopoieticAgents #EncryptedAIMesh #OperationalAntifragility #1to3YearsAhead #QuantumAIDominance #SovereignAI #AIWithGuardrails #NationalSecurityAI #DefenseTechAI #SemiconductorAI #NSTCMember #EthicalGovernanceAI #QuantumBIOS #TechMoat #InvestorReadyAI #QuantumEdge #StrategicDifferentiation #FutureProofAI #DefenseReadyTech #BlackOpsAI #AIForNationalSecurity #GameChangerAI #NoOneMatchesThis

2017 Tesla memo.. communication bottlenecks and NVIDIA ...URGENT NEWS: Perplexity’s new benchmarks cements UnityGrid QPI as the only production grade, multi-agent, ethically-governed quantum-biomimetic AI mesh in the field ..with a 1–3 year lead on everyone else in deployment maturity and a 2 -4 year lead with the Sovereign E8 Kali Red-Team LLMSEC DevOps Engine quantum-agi-consortium-qmagi… Market Reality Check: Competitors are still in rollout, pilot, or niche specialization. We’re already in production-class multi-node, live synchronization, adaptive cycles, and real-time explainability. Innovation Edge: Nobody else combines biomimetic resonance, E8 topology, EchoRhythm Core, and encrypted P2P mesh with dynamic ethical governance. Differentiation Hook: We’re not just building algorithms ...we’ve operationalized autopoietic agents with mission-critical AI guardrails. Strategic Positioning: UnityGrid isn’t chasing qubit counts or LLM compression bragging rights. We own the intersection of quantum AI, national security, and ethical AI governance ...the exact gap where Big Tech has expertise debt and we have a moat. #UnityGridQPI #QuantumPulseIntelligence #E8TopologyResonance #EchoRhythmCore #BiomimeticAI #EthicalAILeadership #MissionCriticalAI #AutopoieticAgents #EncryptedAIMesh #OperationalAntifragility #1to3YearsAhead #QuantumAIDominance #SovereignAI #AIWithGuardrails #NationalSecurityAI #DefenseTechAI #SemiconductorAI #NSTCMember #EthicalGovernanceAI #QuantumBIOS #TechMoat #InvestorReadyAI #QuantumEdge #StrategicDifferentiation #FutureProofAI #DefenseReadyTech #BlackOpsAI #AIForNationalSecurity #GameChangerAI #NoOneMatchesThis

Perplexity’s new benchmarks cements UnityGrid QPI as the only production grade, multi-agent, ethically-governed quantum-biomimetic AI mesh in the field ..with a 1–3 year lead on everyone else in deployment maturity and a 2 -4 year lead with the Sovereign E8 Kali Red-Team LLMSEC DevOps Engine Market Reality Check: Competitors are still in rollout, pilot, or niche specialization. We’re already in production-class multi-node, live synchronization, adaptive cycles, and real-time explainability. Innovation Edge: Nobody else combines biomimetic resonance, E8 topology, EchoRhythm Core, and encrypted P2P mesh with dynamic ethical governance. Differentiation Hook: We’re not just building algorithms ...we’ve operationalized autopoietic agents with mission-critical AI guardrails. Strategic Positioning: UnityGrid isn’t chasing qubit counts or LLM compression bragging rights. We own the intersection of quantum AI, national security, and ethical AI governance ...the exact gap where Big Tech has expertise debt and we have a moat. #UnityGridQPI #QuantumPulseIntelligence #E8TopologyResonance #EchoRhythmCore #BiomimeticAI #EthicalAILeadership #MissionCriticalAI #AutopoieticAgents #EncryptedAIMesh #OperationalAntifragility #1to3YearsAhead #QuantumAIDominance #SovereignAI #AIWithGuardrails #NationalSecurityAI #DefenseTechAI #SemiconductorAI #NSTCMember #EthicalGovernanceAI #QuantumBIOS #TechMoat #InvestorReadyAI #QuantumEdge #StrategicDifferentiation #FutureProofAI #DefenseReadyTech #BlackOpsAI #AIForNationalSecurity #GameChangerAI #NoOneMatchesThis

Are advanced reasoning models easier to jailbreak? LLMs use chain-of-thought reasoning do better on math and coding benchmarks Attackers can trick models into misbehaving using jailbreaks and prompt-based attacks LLMs making Malware and XSS attacks can bring down agentic AI systems they're in We investigated whether adding a reasoning LLM to your agent makes it easier to hijack: 1. Reasoning models are on average MORE robust than non-reasoning models 2. HOWEVER they are 2-3x more likely to fall for Suffix Injection and Tree-of-Attacks with Pruning 3. They're specifically less likely to generate Malware and XSS attacks which means they're less likely to bring down your agent The verdict is that Reasoning models are a great choice for agents! Super grateful to accept the "Best Paper Award" at LLMSEC #acl2025 for this work! Big thanks to my co-authors Aaditya Rastogi and Erick Galinkin as well as the organizers @LeonDerczynski Jekaterina Novikova, PhD, and Muhao Chen

Amazing LLMSEC keynote by @wunderwuzzi23. Yet another call to focus on people - "we need humans to lead AI, not AI to lead humans" #acl2025nlp #acl2025

LLMSEC proceedings are up! sig.llmsecurity.net/proceedi… (Anthology is processing) #ACL2025NLP

That was fast (LLMSEC keynote)

At ACL in Vienna? Hear the world's leading prompt injector talk at LLMSEC on Friday! Johann Rehberger @wunderwuzzi23 will be presenting the afternoon keynote at 14.00 in Hall B > sig.llmsecurity.net/workshop… #ACL2025NLP #ACL2025

Come to LLMSEC at ACL & hear Niloofar's keynote "What does it mean for agentic AI to preserve privacy?" - @niloofar_mire, Meta/CMU (Friday 1st Aug, 11.00; Austria Center Vienna Hall B) See you there! #acl2025 #acl2025nlp

First keynote at LLMSEC 2025, ACL: "A Bunch of Garbage and Hoping: LLMs, Agentic Security, and Where We Go From Here" Erick Galinkin Friday 09.05 Hall B Details: sig.llmsecurity.net/workshop… - #ACL2025NLP

I will be at ACL in Vienna next week. You can hear me at: * Sunday 27th, 1400-1730: Guardrails and Security for LLMs (tutorial) * Monday 28th, 1400-1530: NLP Security and Ethics, in the Wild (presentation, TACL) * Tuesday 29th, 1000-1200: We can't do it alone (industry track keynote) * Wednesday 30th, 1100-1230: NLP Careers panel * Saturday 2nd, all day: LLMSEC 2025 (chairing the workshop) See you around - let's catch up if you're in town!

I will be at ACL in Vienna next week. You can hear me at: * Sunday 27th, 1400-1730: Guardrails and Security for LLMs (tutorial) * Tuesday 29th, 1000-1200: We can't do it alone (industry track keynote) * Wednesday 30th, 1100-1230: NLP Careers panel * Saturday 2nd, all day: LLMSEC 2025 (chairing the workshop) See you around - let's catch up if you're in town! #ACL2025NLP

Excited to give a keynote at the LLMSEC 2025 workshop. The workshop is part of the 63rd Annual Meeting of the Association for Computational Linguistics (ACL) Looking forward to connect more with the academic research community And it's gonna be in Vienna! 🙂

Scaling AI fast? Don’t skip security. LLMSec by Ebryx offers modular, expert-led services to safeguard your LLMs & agents. Protect innovation with a free security check: ebryx.com/llm-sec #AIInnovation #Cybersecurity