AI Security Startups - Top 30 - 2025 https://t.co/ufCZRfJmJn

Adversarial Attacks and Defenses: Deep Reinforcement Learning 1. We propose a classification framework for adversarial attack methods against DRL models based on perturbation types and attack targets. 2. Leveraging this framework, we comprehensively review existing

Attack Strategies for LLM Web Agent Red-Teaming In this work, we presented Genesis, a web agent red-teaming framework that systematically discovers, summarizes, and evolves reusable attack strategies. By introducing the genetic algorithm with a hybrid strategy representation

OpenAI Atlas Omnibox Prompt Injection: URLs That Chat - https://t.co/IKx2YMxE1I - @NeuralTrustAI In OpenAI Atlas, the omnibox (combined address/search bar) interprets input either as a URL to navigate to, or as a natural-language command to the agent. We’ve identified a prompt

Top 6 MCP Vulnerabilities (and How to Fix Them) - https://t.co/yZXw7EOPTB @omercnet at @descopeinc #AIsecurity #MCP #AgentSecurity #PromptInjection #SessionHijacking #ToolingSecurity #Mitigation #ProtocolSecurity #LLM #Hardening #descopeinc #descope

⚠️ Heads up!!! Big warning for HP AI Devices! ⚠️ Some of HP’s latest Next Gen AI PCs, including the EliteBook X Flip G1i, are getting the updated OneAgent 1.2.50.9581 build. That version seems to run a cleanup script removing any certificate containing “1E” in its subject ....

How a fake AI recruiter delivers five staged malware disguised as a dream job #BeaverTail #Lazarus https://t.co/FE8STf5m6H

For Cyber Security Awareness Month at IBM, I delivered an internal cross-org presentation on benchmarking SOTA LLM models using some CTF problems I created this year. I also shared some of my learnings from working extensively on agents 🖤⚔️ Slides => https://t.co/7Y2iYXsrdG

CVE-2025-6515: Prompt Hijacking in MCP ecosystems - https://t.co/b75l2qyi8J by @JFrog, @JFrogSecurity JFrog Security Research recently discovered and disclosed multiple CVEs in oatpp-mcp – the Oat++ framework’s implementation of Anthropic’s Model Context Protocol (MCP) standard.

The New Reality of Threat Modeling in AI-Generated Systems - https://t.co/rqMhmY9agA #ThreatModeling #AIGeneratedCode #SupplyChain #DataPoisoning #OWASPTop10 #MITREATLAS #PromptInjection #ModelIntegrity #SBOM

Factory Recap: Securing AI Agents in Production - https://t.co/khr4XGiQv7 - @GoogleCloudTech #AISecurity #AIAgents #PromptInjection #ContextPoisoning #VectorDBAttack #Sandbox #ModelArmor #RAGLeaks #SupplyChain #EUAIAct

Blog Series: Securing the Future: Protecting AI Workloads in the Enterprise - https://t.co/vzqerjzjFw - @Azure #AI #AISecurity #AISupplyChain #DataPoisoning #ModelBackdoors #MLOps #NIST #MITRE #CloudSecurity #EnterpriseAI

The Highs and Lows of Vibe Coding - https://t.co/8NVAPlXlUh - @snyksec The vibe coding revolution has created billion-dollar companies in months and democratized software creation for millions, while simultaneously introducing catastrophic security vulnerabilities and

Hugging Face and VirusTotal collaborate to strengthen AI security - https://t.co/ICxMhv0CkQ - @huggingface @bquintero @XciD_ Starting today, every one of the 2.2M+ public model and datasets repositories on the Hugging Face Hub is being continuously scanned with VirusTotal. Why

Interpreting Jailbreaks and Prompt Injections with Attribution Graphs - https://t.co/31fulJ9D4C by @zenitysec Today’s agent security is strong at the edges: we monitor inputs/outputs, trace and permission tool calls, track taint, rate-limit, and log everything. We have a very

Leveraging Machine Learning to Enhance Acoustic Eavesdropping Attacks (Blog Series) - https://t.co/fY6ESiEqBq In this post, we discussed the various steps of data preprocessing we took to prepare the accelerometer and gyroscope readings for AI training. We also discussed the

1/ NEW: We propose a new black-box attack on LLMs that needs only text (no logits, no extra models). It's generic: we can craft adversarial examples, prompt injections, and jailbreaks using the model itself👇 How? Just ask the model for optimization advice! 🎯

Metanarrative Prompt Injection - https://t.co/Of8awtC6O3 by @rez0__ When exploiting AI applications, I find myself using this technique really often so I figured I’d write a quick blog about it. I call it the “Metanarrative Prompt Injection.” You might have already used this