> Markdown version of [/videos/2152-how-to-stop-your-agents-from-going-rogue-arnav-gupta](https://www.wearedevelopers.com/videos/2152-how-to-stop-your-agents-from-going-rogue-arnav-gupta). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# How to Stop Your Agents From Going Rogue - Arnav Gupta

Arnav Gupta warns that deploying autonomous AI agents without guardrails is a disaster waiting to happen. Learn how open-source runtime security stops rogue systems from deleting your production data.

- **Speakers:** 
- **Event:** 
- **Published:** August 21, 2026
- **Duration:** 24:09
- **URL:** https://www.wearedevelopers.com/videos/2152-how-to-stop-your-agents-from-going-rogue-arnav-gupta

## Summary

As organizations rush to deploy autonomous AI agents as virtual coworkers, they inadvertently grant these systems highly privileged access without adequate guardrails—akin to driving a car without brakes. To solve this critical vulnerability, Arnav Gupta founded Prismor, an open-source runtime security layer that monitors and blocks suspicious agent actions. By embedding business logic and requiring a "human in the loop" for high-stakes decisions, Prismor prevents scenarios where rogue agents could execute disastrous commands, such as deleting production databases or making highly unethical automated trades.

The transition to fully autonomous, multi-agent architectures has fundamentally shifted cybersecurity from a deterministic discipline to a probabilistic one, driven by the infinite ways an LLM can be prompted. While individual developers often view security as a productivity hurdle under the "move fast and break things" mantra, enterprises face severe compliance pressures, such as the EU AI Act, which threatens fines up to 2% of annual revenue. Because of these stakes, highly regulated sectors like fintech and healthcare are driving the demand for agent guardrails. In these industries, agents must not only achieve objective goals but also understand the subjective, ethical context of their actions to protect brand reputation and sensitive user data.

Rather than maintaining a closed ecosystem, Prismor relies on an open-source strategy to serve as a signifier of trust, transparency, and long-term commitment to enterprise clients. By allowing the community to vet the underlying security algorithms, the platform mitigates fears of supply chain vulnerabilities and vendor abandonment. Ultimately, robust agent security acts as an enabler rather than a roadblock, curing enterprise AI adoption FOMO by allowing companies to safely deploy cutting-edge technologies while relying on the collaborative power of open-source development.

**Keywords:** AI agent runtime security, open-source security frameworks, autonomous agent guardrails, probabilistic cybersecurity, EU AI act compliance, enterprise AI adoption blockers, human-in-the-loop validation, fintech AI regulations, healthcare data protection, multi-agent architecture risks, LLM prompt vulnerabilities, automated trading security, agentic retrieval systems, vendor lock-in mitigation, supply chain attack prevention

## Chapters

1. **Moving from India to San Francisco for startup accelerators** (00:02) — Immersing in the local builder culture accelerates the journey of creating new software and robotics companies.
1. **Runtime security layers for autonomous artificial intelligence agents** (02:03) — Implementing a human-in-the-loop framework prevents autonomous agents from executing suspicious actions against business logic.
1. **Treating autonomous agents as privileged identities requiring safety guardrails** (04:00) — Deploying artificial intelligence with full access necessitates strict guardrails to prevent infrastructure damage and security breaches.
1. **Balancing rapid artificial intelligence development with strict compliance regulations** (05:56) — Integrating security guardrails protects critical user data and ensures compliance with frameworks like the EU AI Act.
1. **Transitioning from open source retrieval systems to enterprise security** (09:06) — Witnessing rogue agents destroy production databases highlights the critical need for context-aware safety systems in enterprise deployments.
1. **Utilizing open source models to establish trust with enterprises** (13:12) — Maintaining a transparent codebase allows enterprise clients to verify security algorithms before integrating third-party tools.
1. **Defending against probabilistic vulnerabilities in nascent agent ecosystems** (16:43) — Developing comprehensive platforms to auto-fix vulnerabilities mitigates risks associated with infinite prompting and supply chain attacks.
1. **Overcoming data privacy concerns to safely adopt artificial intelligence** (19:39) — Implementing robust security frameworks enables cautious enterprises to utilize artificial intelligence without violating regional data privacy laws.
1. **Collaborating on secure coding environments and open sourcing solutions** (21:29) — Partnering with infrastructure creators to secure virtual cloud environments demonstrates the value of community collaboration and open sourcing tools.

## Related Moments

- [From read-only models to excessive agency](https://www.wearedevelopers.com/videos/100038-the-day-the-chatbot-asked-for-sudo) (from "The day the chatbot asked for sudo")
- [Security integration and AI skepticism in developer tooling](https://www.wearedevelopers.com/videos/1830-wearedevelopers-live-speculaitions) (from "WeAreDevelopers LIVE - SpeculAItions")
- [Current state of security in AI applications](https://www.wearedevelopers.com/videos/1637-delay-the-ai-overlords-how-oauth-and-openfga-can-keep-your-ai-agents-from-going-rogue) (from "Delay the AI Overlords: How OAuth and OpenFGA Can Keep Your AI Agents from Going Rogue")
- [Managing security risks introduced by autonomous AI agents](https://www.wearedevelopers.com/videos/1403-five-things-in-tech-that-matter-and-we-have-to-make-work) (from "Five things in tech that matter and we have to make work")
- [Designing an open agent platform with governed autonomy](https://www.wearedevelopers.com/videos/1538-composable-intelligence-how-henkel-and-microsoft-are-shaping-the-agent-ecosystem) (from "Composable Intelligence: How Henkel and Microsoft Are Shaping the Agent Ecosystem")
- [Fusing developer experience and platform engineering for agentic SDLC](https://www.wearedevelopers.com/videos/100266-ai-won-t-fix-your-engineering-culture) (from "AI Won't Fix Your Engineering Culture")

## Related Articles

- [What is Agentic Programming and Why Should Developers Care?](https://www.wearedevelopers.com/magazine/625-what-is-agentic-programming-and-why-should-developers-care)
- [Stephan Gillich - Bringing AI Everywhere](https://www.wearedevelopers.com/magazine/489-stephan-gillich-bringing-ai-everywhere)
- [Trustworthy AI Starts at Deployment: 5 Checks Before You Ship](https://www.wearedevelopers.com/magazine/753-trustworthy-ai-starts-at-deployment-5-checks-before-you-ship)
- [Never delegate the understanding](https://www.wearedevelopers.com/magazine/749-never-delegate-the-understanding)

## Related Jobs

- [Security Architect - AI](https://www.wearedevelopers.com/jobs/ext/1581899-security-architect-ai) at **ZEISS Group**
- [Senior AI/ML Engineer](https://www.wearedevelopers.com/jobs/48352-senior-ai-ml-engineer) at **PagerDuty**
- [Senior AI Agent Software Engineer (Go, Python) (m/f/x)](https://www.wearedevelopers.com/jobs/48277-senior-ai-agent-software-engineer-go-python-m-f-x) at **Dynatrace**
- [Principal Software Engineer, Enterprise AI Platform](https://www.wearedevelopers.com/jobs/ext/1467292-principal-software-engineer-enterprise-ai-platform) at **GitHub**
- [Senior Product Manager](https://www.wearedevelopers.com/jobs/48353-senior-product-manager) at **PagerDuty**
- [AI Software Engineer (Germany)](https://www.wearedevelopers.com/jobs/48317-ai-software-engineer-germany) at **Sunhat**