Andromeda Security

We're #hiring a new Frontend/UI Developer in Bengaluru, Karnataka. Apply today or share this post with your network.

We're #hiring a new Member of Technical Staff in Bengaluru, Karnataka. Apply today or share this post with your network.

Monotosh Das / Sauvik Biswas are hiring. The team works on cutting edge AI/Identity security domain and requires backend developers who can be a force multiplier with AI tools. To know more about andromeda security visit Andromeda Security page on linkedin. To apply, pls ensure you fall in the exp bracket of 4+ years and have expert distributed systems knowledge. Experience of working in a cloud native/ai native software company is a big plus. Pls send your resume to monotosh@andromedasecurity.com or sauvik@andromedasecurity.com

Governing agents with just tool calls or OAuth scopes is too coarse and misses the actual picture of what the agent touches. Most of these agents are getting built in quick time and while governance is getting built at tool call level, it is not enough. At Andromeda Security we bring our access graph into our policy enforcement engine to build unprecedented granular controls by leveraging our existing work in this space. Pls checkout the link below that discusses about this topic in detail: https://lnkd.in/g2TiM_ME Would love to hear your thoughts. Murali Basavaiah Ashish Shah

🚀 Big news from the Andromeda Security team today! We are proud to announce a major expansion of our unified identity security platform, bringing active, Runtime Agentic AI enforcement to the enterprise. Passive visibility isn't enough to secure autonomous agents. This release bridges the agent enforcement gap by delivering real-time, resource-level controls alongside full-lifecycle JML capabilities, all managed through an intent-driven conversational interface. Enterprise security leaders are already putting it to work: "Scaling AI safely demands active agent security, not passive monitoring... unifying them with human JML capabilities into a single continuous compliance engine is what sets this apart." - Preston Horne at Dark Matter Technologies "Andromeda’s inline AI access management addresses the agent enforcement gap... providing us with the holistic context and inter-relationships across agents, NHIs, and humans needed to scale safely." - Bill Harper at New American Funding 👉 Read the full press release: https://lnkd.in/e6CRrups 👉 Explore the runtime enforcement use case: https://lnkd.in/eMdVWzTk Are you at Identiverse 2026 in Las Vegas this week? Stop by the Andromeda booth 317 to see a live demonstration of these new capabilities in action! #AISecurity #IdentityGovernance #AgenticAI #Identiverse2026

Your AI agent’s real permissions aren’t the ones you granted it. When security teams onboard an agent, they define its upper bounds—what tools and resources it may use. But real capability is a different story. If your agent runs on a shared identity or overly broad OAuth scopes, it can easily cross a dangerous line. When the target application can no longer see the end-user's individual ceiling, you haven’t built a helpful assistant—you’ve built a privilege-escalation machine with a friendly chat interface. In cybersecurity, this is the confused deputy problem, and it’s the dominant risk in agent deployments today. To stop it, security must enforce a strict three-way intersection: 1. Agent Allowances: What the agent is permitted to touch. 2. Token Scopes: What the OAuth scope actually grants. 3. User Entitlements: What the human in the loop is genuinely allowed to do. Because target applications go blind in shared-identity models, the gateway layer is the only architectural point that can enforce all three simultaneously. Check out our latest deep dive at Andromeda to see how we are building agentic security that respects the real ground truth of human identity. 👇 Read the full blog here https://lnkd.in/g4i3YugH #AIsecurity #GenAI #LLMOps #Cybersecurity #IdentityAccessManagement

It’s easy to give an AI agent access to an enterprise tool. It’s much harder to control what it actually does once it’s inside. Many teams rely on MCP to manage agent tools. MCP is fantastic for tool discovery, but it has a massive blind spot: it looks at the tool level, not at the underlying business-critical assets. Think about an agent calling an execute_sql tool. The token is valid and the tool is approved. But the protocol layer doesn't know the difference between querying sandbox data or dropping a production database. As we move agents into workflows that matter—like database migrations and customer refunds—the failure mode is no longer a bad summary. It’s an outage or a data leak. We must shift our focus from "Can this agent use this tool?" to "Can this agent touch this specific business-critical asset right now?" Sudipto Biswas broke down how we are bridging this gap at Andromeda Security in this latest post: https://lnkd.in/gyxSgh9T #AIAgents #ProductManagement #CyberSecurity #MCP #AndromedaSecurity

The shift to autonomous AI agents is forcing a massive rethink of infrastructure security. Are your current OAuth flows ready for headless, ephemeral agents? Traditional Machine-to-Machine (M2M) authentication breaks down when agents dynamically decide which tools to call. Relying on standard OAuth scopes alone leads to token bloat and a dangerous lack of granularity. In our latest blog, Securing the New Perimeter: AI Gateways, we explore how to secure this new frontier. https://lnkd.in/gDQ7ugdw Highlights include: * The Agent Dilemma: Why static credentials fail for thousands of short-lived, task-specific agents. * The AI Gateway: How deploying a zero-trust enforcement point solves complex auth challenges. * Delegated Access: Maintaining audit trails and provenance when agents act on behalf of human users. Read the full breakdown and see how Andromeda is bridging the gap between strict policy governance and modern authentication: https://lnkd.in/gDQ7ugdw #Cybersecurity #AIAgents #ZeroTrust #OAuth #IdentityManagement #Andromeda

As AI agents become deeply integrated into our daily workflows and enterprise applications, traditional security boundaries are rapidly evolving. We are no longer just protecting the network—the new defense frontier is the Application Perimeter. 🛡️🤖 At Andromeda, we know that securing this new perimeter means treating Agentic AI as a full identity class. Autonomous agents interacting directly with APIs and external data require the exact same rigorous Access Intelligence, least-privilege management, and continuous governance that we demand for human identities. Our very own Sudipto Biswas just published an excellent deep dive on exactly how this landscape is shifting. A few high-level takeaways: 🔹 The Shifting Perimeter: Why traditional network defenses fall short when autonomous agents act on behalf of users. 🔹 Access & Guardrails: The absolute necessity of strict authorization, input validation, and rightsizing permissions for AI tools. 🔹 New Attack Surfaces: Navigating and mitigating emerging threats like indirect prompt injections and unauthorized agent actions. If you are building, deploying, or managing AI agents, understanding this new perimeter is critical to preventing compromised agents and shadow AI. 👇 🔗 Read Sudipto’s full piece here: https://lnkd.in/gpVWjNJ2 #AIAgents #Cybersecurity #IdentitySecurity #ApplicationSecurity #AndromedaSecurity #NHI #ArtificialIntelligence #LeastPrivilege

Excited to close out another incredible quarter at Andromeda Security. We’re continuing to push the boundaries of what’s possible in identity security—innovating, disrupting, and helping organizations stay ahead of an increasingly complex threat landscape. The time is now, and we’re leaning in. Proud to share that we grew the business 50% this quarter. Results like that don’t happen by accident—they’re driven by relentless execution, strong partnerships, and a clear vision. Couldn’t be prouder of this team and everything we’re building together. Onward 🚀 Murali Basavaiah Ashish Shah Daniel Weaver Mark T. Ian Forrest Patrick O'Brien Talton Taylor