Prompt injection used to be a problem about a single input. Defenders inspected a string, or an image, or an audio clip, and asked whether that single channel carried a malicious instruction. Multimodal models break that assumption.
Security research and technical insights
Security in the Age of AI
NVIDIA advances AI security by helping defenders apply AI and helping organizations create safer AI systems.
Research and technical insights
How does NVIDIA approach AI security?
NVIDIA helps shift the asymmetric advantage toward defenders by sharing frontier security research openly, enabling the community to pool knowledge, resources, and expertise.
Transparent Insights
Make AI security research open to inspection, adaptation, and validation across models, to enable stronger protections across the full AI stack.
Reproducible Results
Develop repeatable methods, clear benchmarks, and research artifacts so teams can validate findings and build on them.
Practical Security
Produce actionable security research to drive vulnerability discovery, disclosure, remediation, and prevention to help defenders manage real-world risk.
Published work
AI security contributions
The growing demand for accelerated computing is driving widespread deployment of multi-accelerator systems in the cloud and at the edge.
Large Language Models (LLMs) have enabled the development of powerful agentic systems capable of automating complex workflows across various fields.
Large reasoning models (LRMs) extend large language models with explicit multi-step reasoning traces, but this capability introduces a new class of prompt-induced inference-time denial-of-service (PI-DoS) attacks that exploit the high computational co
AI agents, predominantly powered by large language models (LLMs), are vulnerable to indirect prompt injection, in which malicious instructions embedded in untrusted data can trigger dangerous agent actions.
Privacy protection has become an increasing concern in modern machine learning applications.
Private information retrieval (PIR) allows private database queries; however, it is hindered by intense server-side computation and memory traffic.
Multi-user virtual reality enables immersive interaction. However, rendering avatars for numerous participants on each headset incurs prohibitive computational overhead, limiting scalability.
Approximate nearest neighbor (ANN) search in AI systems increasingly handles sensitive data on third-party infrastructure.
While private information retrieval (PIR) enables private database services by fully concealing access patterns, it simultaneously requires high computational throughput, large memory capacity, and substantial memory bandwidth.
Creating an asymmetric advantage through community
The Open Secure AI Alliance unites a community of security experts working together to keep the world safe.
FAQs
Why does NVIDIA share AI security research openly?
AI security is a collective challenge. NVIDIA shares research, methods, and technical artifacts so defenders can inspect findings, build on them, and help strengthen protections across the ecosystem.
NVIDIA is an active contributor to the Open Secure AI Alliance, managed by the Linux Foundation, which provides a community for security teams to collaborate on this work. Learn more.
What does NVIDIA contribute to AI security?
NVIDIA AI security research studies how AI systems can be secured and how AI can help defenders. The work spans open systems, reproducible cyber evaluation, red teaming, runtime controls, human accountability, plus vulnerability discovery, reporting, and remediation.
What security topics does this page cover?
This collection covers AI agent security, prompt injection, offensive and defensive cybersecurity, vulnerability discovery and remediation, model-weight security, safety evaluation, multimodal attacks, observability, and governed deployment. It also connects this research to real-world cyber defense and secure AI infrastructure.
What makes AI security insights useful to practitioners?
Useful research defines the security problem, documents the method, states limitations, and provides evidence others can inspect, test, or reproduce. NVIDIA AI Security prioritizes practical work that helps teams evaluate AI systems and apply findings to stronger defenses.
Where can I find published NVIDIA Research papers?
Find published papers in the NVIDIA Research publications directory. Individual security-research entries should also link directly to their paper, technical blog, code, data, or other public primary source.
Where can I find information about NVIDIA technologies for AI security?
Explore NVIDIA cybersecurity AI technologies, including open AI models and libraries, secure infrastructure, and reference examples.