Robust Intelligence: Securing the AI Transformation
Robust Intelligence, now a Cisco company, provides an AI Application Security Platform that helps enterprises achieve AI security and safety. Their platform addresses the challenges of securing AI applications, which often involve multiple stakeholders, diverse tools, and limited visibility for security teams.
The Problem: Securing AI in Enterprises
The development and deployment of AI-powered applications introduce new security and safety risks. These risks are often spread across different teams, requiring specialized expertise and collaboration. The decentralized nature of AI teams, the variety of tools used, and the rapid evolution of AI technology further complicate the issue.
The Solution: End-to-End Security for AI Applications
Robust Intelligence offers a comprehensive solution with two main components:
- AI Validation: Automates the evaluation of AI models, data, and files for security and safety vulnerabilities. It helps determine necessary guardrails for secure AI deployment in production.
- AI Protection: Provides guardrails for AI applications in production, protecting against integrity, privacy, abuse, and availability violations. It uses automated threat intelligence platform updates.
Key Benefits:
- Unblock the enterprise AI mission: Deploy applications in minutes instead of months or years.
- Decouple AI development from AI security: Save time and resources by automating security processes.
- Automate AI security excellence: Meet industry standards (NIST, MITRE ATLAS, OWASP LLM Top 10) with a single integration.
- Align AI security across stakeholders: Improve collaboration between AI, security, and compliance teams.
- Protect against evolving threats: Identify novel, zero-day vulnerabilities and block bad actors.
Technology Behind the Platform
Robust Intelligence's platform uses proprietary technology developed over a decade, combining algorithmic red teaming, threat intelligence pipelines, and policy mappings. This creates a model engine that generates examples of inputs to expose vulnerabilities, continuously improving the AI Validation and AI Protection products.
Robust Intelligence's Role in Shaping AI Security Standards
Robust Intelligence has been a key contributor to the AI security landscape, co-developing the AI Risk Database, co-authoring the NIST Adversarial Machine Learning Taxonomy, and contributing to the OWASP Top 10 for LLM Applications.
Recognized Achievements:
- Demonstrated the impact of data poisoning on large language models.
- Identified security vulnerabilities in NVIDIA's NeMo Guardrails.
- Achieved an algorithmic jailbreak of GPT-4 and Llama-2.
- Received the ICML Test of Time Award for work on data poisoning.
Partnerships
Robust Intelligence partners with various organizations to deliver comprehensive AI security solutions.
Resources and Insights
Robust Intelligence provides valuable resources, including articles on algorithmic AI red teaming, Llama-2 risk assessment, and the NIST Adversarial ML Taxonomy.
Conclusion
Robust Intelligence offers a robust and comprehensive approach to securing AI applications, helping enterprises navigate the complexities of AI security and safely deploy AI-powered solutions.
