Latio's 5 key takeaways from Black Hat and Defcon

🆕 New month. New software categories. From AI SOC assistants defending your security ops to contact center AI observability keeping your AI agents in check — August’s 7 new categories reflect where tech is headed. Explore the latest additions, top-rated products, + what they mean for your business at the 🔗 below!

AI is moving faster than security can keep up. Shadow models, unchecked third-party tools, and sensitive data leakage create blind spots and regulatory risk. Join Nimrod Vax (BigID) & Christopher Steffen (Enterprise Management Associates (EMA)) as they unpack how CISOs can bring order to AI chaos: catalog models, enforce policies, monitor usage, and align with governance before regulators mandate it. 👉 Register now: https://bit.ly/46jyUeq Hosted by: Information Systems Security Association (ISSA)

Missing the "semester" of practical AI security increases exposure to prompt injection, data leakage, and supply‑chain risk. OWASP Top 10 for LLM provides a taxonomy of common vulnerabilities that should inform threat models. MITRE ATLAS complements that taxonomy by mapping attacker methodologies and supporting red teaming exercises. Short technical priorities: perform input/output handling tests, validate model supply chains, test for training data poisoning, and constrain model agency. Examples of tooling noted in field work include Garak, Promptfoo, Llama Guard, and NVIDIA NeMo Guardrails but tool choice should follow organizational risk requirements. Operational recommendation: align red teaming with business-critical workflows, log model interactions, and introduce guardrails and monitoring before production rollouts. 🚨 #LLMSecurity #MITREATLAS #OWASP https://lnkd.in/dmw8yJVw

When integrating LLMs with private data, security can be a serious hurdle. But it doesn't have to be. 🔐 Snowflake Cortex AI addresses essential security must-haves, from authentication and access controls to encryption and compliance. This frees up developers to focus on what matters most: building applications with leading models from Anthropic, OpenAI, Mistral, DeepSeek, and Meta. Learn how to build with confidence. 👉 https://bit.ly/4mNkA4M

Exciting & ground-breaking news from BeyondTrust today! We’re bringing industry-first, production-ready identity security controls for agentic AI into Identity Security Insights and the BeyondTrust Pathfinder Platform. The new capabilities will deliver: - Comprehensive AI Agent Insights - MCP Orchestration with Lightweight MCP Server - Omnipresent AI Decision-Support Layer This means customers will get: - Full visibility across human, machine, secrets & AI identities - Faster remediation with real-time intelligence - Unified governance & enforcement Full details here: https://lnkd.in/dbKdyTiA

CISOs don’t just need visibility—they need credibility. Agentic AI isn’t just speeding up detection; it’s helping security leaders prove how their SOC reduces risk and drives measurable outcomes. Exabeam Nova delivers posture summaries, ATT&CK coverage benchmarking, and reports that speak board language—not just tool telemetry. For anyone working to elevate security’s role in the business, this one’s for you: https://ow.ly/LqiX30sP911

Real-Time Network Cost Intelligence with Mav Turner The interview with Mav Turner, Chief Product Officer at Kentik, examines the evolution of network security and the critical role of network intelligence. Mav shares insights from his career and discusses how Kentik's platform provides visibility into network traffic to optimise performance and cost. The conversation also explores the impact of AI and machine learning on network management and the future of data centre operations. Watch on YouTube: https://lnkd.in/gkr6WhMQ Jordi Mon Companys | The Software With

Glad to share a new post on AI Security, where I conceptualize a new term ContextSecOps (Context Guardrail). I begin by Embedding LLM Backdoor to allow Jailbreak, move to 2 Multi Agent System (RED/BLUE) having a bottle (prompts and implementation using Google ADK included), and I finalize with MCP ContextSecOps & ToolSecOps. Many Cool resources and references are included. https://lnkd.in/du-AJFDy

Is AI a friend or foe? Invicti CEO, Neil Roseman, dives deep into this critical question in a recent article in Computing Security Magazine Discover his insights on the impact of AI on security and how we can harness its power responsibly. Read the article, and join the conversation! 🔗 https://okt.to/9om1BC

Sometimes when you really just want your AI agent to execute a task, just send it an email. Not only will it execute the task, it’ll give you a detailed response of what was done. Think about all those SOC/NOC analysts dealing with alert fatigue. Just send all alerts to the AI agent and let it sort them out autonomously. That’s why we created CyberAgent - our human computer interaction system. It also allows agent-to-agent messaging too.