Amazon partnership establishes Cornell AI security initiative
Amazon partnership establishes Cornell AI security initiative
Publish Date: 2026-06-10 11:06:00
Source Domain: news.cornell.edu
- Cornell computer scientists, with support from Amazon, are developing safety protocols for AI agents to mitigate the risk of generating harmful code.
- A new Cornell initiative, AI4AI, is being established to improve agentic AI through collaboration among faculty in machine learning, security, formal methods, and verification.
- The project, led by professors Alexandra Silva and Vitaly Shmatikov, aims to explore how to benefit from AI agents’ potential in software development while ensuring code integrity and security.
- AI agents, currently capable of writing software from prompts, lack skepticism and can produce code with security weaknesses that can be exploited.
- The Cornell team intends to create a security framework within AI agents to more cautiously produce code by incorporating rules and verification checks.
- Industry-academia collaboration is highlighted as essential for creating robust security tools and frameworks for AI code generation.
