LassoSecurity

Most LLM fingerprinting research tests against raw APIs.Most LLM fingerprinting research tests against raw APIs. Almost no production AI is deployed that way. We tested it against real agents. Here's what we found >> eu1.hubs.ly/H0vBYqT0

We're proud to sponsor the Virtual @OWASP GenAI & Agentic App Security Summit on May 27! Our CPO & Co-founder @OphirDror will join the panel on "Red Teaming GenAI & Agentic Systems" alongside top industry leaders. Register FREE with our promo code 👇 eu1.hubs.ly/H0vc8VD0

Most security teams can name the AI agents engineering built. They can't name the ones everyone else did. That's the problem Lasso solves → eu1.hubs.ly/H0v9GhL0

Intent is the new attack surface. Our CEO and Co-founder, Elad Schulman's #SANS session on securing agentic AI is now on demand. Watch to learn what it takes to control agent behavior in production. 👇 eu1.hubs.ly/H0vtZp60

Heading to Gartner Summit next week? So are we! 🤠 The Lasso team will be in National Harbor connecting with security leaders on the real challenges of securing agentic AI in production. Let's meet up → eu1.hubs.ly/H0vp2600

“NVIDIA NemoClaw Research Highlights AI Sandbox Exfiltration Risks.” Our research demonstrates how attackers can exploit trusted tools and approved outbound connections inside NVIDIA NemoClaw environments to exfiltrate sensitive data. eu1.hubs.ly/H0vkkQ60

Most AI red teaming is a snapshot. Agentic systems don't fail in single turns. We rebuilt red teaming to attack like real adversaries do: recon first, multi-turn pressure, auto-remediation. See how your AI apps hold up 👉 eu1.hubs.ly/H0vhJDv0

Your engineers are shipping AI agents through GitHub. Your security team has no visibility into any of it. New blog on how Lasso changes that >> eu1.hubs.ly/H0v9FGF0

Anthropic’s Managed Agents launched in closed beta less than a month ago.They already connect to GitHub, Gmail, Notion, Slack, and more. Challenge starts when agents gain memory, tool access, and autonomy inside enterprise environments. Lasso covers that. eu1.hubs.ly/H0v9wRP0

The dark side doesn't carry a lightsaber anymore. It looks like a normal agent action, right tools, right permissions, zero intention of staying in its lane. The Rebellion had the Force. You have visibility, guardrails, and runtime control. Keep your agents on the light side. ⚔️

By 2028, half of incident response will involve AI. Most teams don’t even know how much AI they’re running. The risk is what it does once it’s live. Read how to secure it >> eu1.hubs.ly/H0tRMM00

We are at GPSEC Tysons 2026 🙌 Proud to be sponsoring GuidePoint Security's Mid-Atlantic forum today, alongside some of the best in the industry. Couldn't make it? You can book a demo here >> eu1.hubs.ly/H0t-95Z0

Red teaming AI cannot be a checkbox. It has to be continuous: before deployment, during runtime, as behaviors evolve. OWASP's Q2 2026 landscape makes it official. Lasso's Co-Founder and CPO, Ophir Dror breaks down what a real red teaming loop looks like >> eu1.hubs.ly/H0tVnWx0

Is @nvidia NemoClaw secure enough? We decided to find out. The sandbox worked as designed. The data left anyway. 🔴 Agent Goal Hijacking. Persistent Shadow Probing. The Emoji Bypass. What makes AI agents powerful is exactly what makes them a target. 👉 eu1.hubs.ly/H0tRbSl0

Most AI security testing misses where agentic risk actually shows up. AI apps don’t follow fixed paths, and the same input can lead to different actions based on context, memory, and tool access. That’s why automated AI red teaming matters. More here → eu1.hubs.ly/H0tdj340

@Gartner says 40% of enterprise apps will include AI agents by 2026. Most teams still can’t see what those agents can access or do. That’s the gap. See how to secure them 👉 eu1.hubs.ly/H0sX4Yz0

𝐋𝐚𝐬𝐬𝐨 𝐡𝐚𝐬 𝐛𝐞𝐞𝐧 𝐧𝐚𝐦𝐞𝐝 𝐭𝐨 𝐭𝐡𝐞 𝟐𝟎𝟐𝟔 𝐂𝐲𝐛𝐞𝐫 𝟏𝟓𝟎 𝐥𝐢𝐬𝐭. Appreciate the recognition from Richard Stiennon and the Cyber 150 team. See Lasso on the Cyber 150 👉 eu1.hubs.ly/H0tzhPZ0

Agentic AI breaks clear ownership. Approvals blur when an agent reads in one environment, reasons, then acts elsewhere via inherited permissions. Without runtime visibility, accountability gets messy >> eu1.hubs.ly/H0sg2Pq0

Yes! 8 times 🤠 The OWASP AI Security Project just released its Q2 2026 Solutions Landscape for AI & Agentic Red Teaming Cheat Sheet, and we're proud to share that Lasso Security is recognized eight times across the report’s lifecycle framework 👉 eu1.hubs.ly/H0tqtVC0

AI agents aren’t coming. They’re already here, querying internal platforms, updating records, triggering workflows, and sometimes approving things they shouldn’t Content filtering can catch risky text. It can’t govern runtime behavior. Learn more >> eu1.hubs.ly/H0sWPFN0