Aardvark: When AI Starts Hunting Its Own Bugs
Aardvark: OpenAI's Autonomous Security Researcher
Author: Mamta Upadhyay
![[Thought Post] Protecting Our Minds in the Age of AI](https://mamtaupadhyay.com/wp-content/uploads/2025/10/steve-johnson-_0iV9LmPDn0-unsplash.jpg)
[Thought Post] Protecting Our Minds in the Age of AI
Rethinking Cognition: Navigating AI’s Influence
![[Threat Model] AI-Augmented Developer Tools](https://mamtaupadhyay.com/wp-content/uploads/2025/10/Screenshot-2025-10-14-at-12.51.18-PM.png)
[Threat Model] AI-Augmented Developer Tools
AI is not just changing how we write code: it is also changing how we secure the act of writing…
The Rise of Agentic Commerce
The next frontier of payment security isn’t in code or crypto, but in the language our models use
![[Thought Post] Exponential AI, Linear Security](https://mamtaupadhyay.com/wp-content/uploads/2025/10/Screenshot-2025-10-04-at-11.10.44-AM.png)
[Thought Post] Exponential AI, Linear Security
AI is evolving exponentially, but our defenses are still built for a slower world
![[Threat Model] Agentic AI](https://mamtaupadhyay.com/wp-content/uploads/2025/09/Screenshot-2025-09-27-at-9.35.44-AM.png)
![[Thought Post] Added Responsibilities of AppSec Leads in AI-First Orgs](https://mamtaupadhyay.com/wp-content/uploads/2025/09/Screenshot-2025-09-13-at-1.20.02-PM.png)
[Thought Post] Added Responsibilities of AppSec Leads in AI-First Orgs
In AI-first organizations, AppSec leadership is no longer just about securing code, it’s about securing the intelligence that drives business.
![[Threat Model] RAG Systems](https://mamtaupadhyay.com/wp-content/uploads/2025/09/ChatGPT-Image-Sep-11-2025-09_40_41-PM.png)
![[Threat Model] AI Memory Systems](https://mamtaupadhyay.com/wp-content/uploads/2025/09/Create-a-highly-detailed-and-sharp-focused-featured-image-for-a-blog-post.-The-image-must-say-the-exact-words-_Threat-Model-AI-Memory-Systems_.jpg)
Memory Collisions in Multi-Agent Systems
When multiple AI agents share memory, a single poisoned entry can silently ripple across the system, reshaping how every agent…
Invisible Inputs in AI Agents
Invisible inputs hide in AI memory as metadata and embeddings, silently steering agents long after the attacker is gone.
Temporal Attacks on AI Memory: Beyond Side-Channels
Temporal attacks exploit recency bias, letting new inputs override trusted memory
Autonomy Escalation in MCP Agents
Autonomous MCP agents can quietly expand their operational scope, turning harmless requests into high-impact actions through a hidden process of…