Assured Integrity for AI-Based Software
How can we trust AI-powered software not to fail, misbehave, or be exploited?
How can we trust AI-powered software not to fail, misbehave, or be exploited?
As artificial intelligence takes on coding, software management, and decision-making with minimal oversight, Cornell researchers are launching an initiative uniting experts in AI, security, and verification to address the risks of autonomous systems that plan, reason, and act.
This new effort brings together experts in artificial intelligence, computer security, and formal verification to tackle the growing risks posed by autonomous AI systems — software agents that don’t just generate text, but plan, reason, and take action in the world.
“AI is no longer just assisting software development, in many cases it is the software. This research is about making sure AI systems act not just intelligently, but responsibly, safely, and in ways we can verify and trust.”
Alexandra Silva, Vitaly Shmatikov
AI4AI Co-Lead Principal Investigtaors
AI4AI Research Team
Alexandra Silva
Principal Investigator
Professor of Computer Science, Cornell Bowers
Works in formal methods, programming languages, and automated reasoning with applications in networking and probabilistic reasoning. Her research includes foundational work on NetKAT, a formally verified programming language and algebra for software‑defined networking that enables rigorous reasoning about network behavior.
Vitaly Shmatikov
Principal Investigator
Professor of Computer Science, Cornell Tech
Works on digital privacy and secure systems. His recent research focuses on vulnerabilities in AI models and systems, including membership inference, poisoning, adversarial influence, and security issues in multi-agent systems, as well as principled protection methods.
Saikat Dutta
Assistant Professor of Computer Science, Cornell Bowers
Works at the intersection of software engineering and machine learning. His research develops techniques for automated testing and debugging to improve the reliability of machine learning-based systems, as well as leveraging machine learning to solve tasks in software engineering.
Greg Morrisett
Jack and Rilla Neafsey Dean and Vice Provost, Cornell Tech
Contributes to formal methods and secure systems, including hardware–software co-design and compilers. He also develops programming language technologies for building secure, reliable, and high‑performance software systems, with major contributions such as typed assembly language and software fault isolation.
Andrew Myers
Professor of Computer Science, Cornell Bowers
Develops expressive programming abstractions that simplify building secure, trustworthy, and scalable software systems. His research bridges programming languages, computer security, and distributed systems, emphasizing language-based methods to ensure strong security across local and distributed computation.
Kevin Ellis
Assistant Professor of Computer Science, Cornell Bowers
Leads research in artificial intelligence (AI) and program synthesis, studying combinations of learning and reasoning that draw on insights from both machine learning and cognitive science.
Rachee Singh
Assistant Professor of Computer Science, Cornell Bowers
Works in systems and networking, developing algorithms and systems for efficient communication over server‑scale, rack‑scale, and long‑haul photonic interconnects. Her research improves the performance of distributed machine learning and large‑scale cloud workloads.
Fred Schneider
Samuel B. Eckert Professor of Computer Science
Works in trustworthy computing---systems that operate correctly despite attacks and failures. His research spans applied and foundational topics in cyber-security, formal methods, and fault-tolerance.
Cornell is shaping the future of AI
Cornell is driving the future of AI through its unparalleled breadth of expertise and leadership in research and education.
The Cornell AI Initiative, a university-wide effort, advances AI as a transformative tool across disciplines—from classrooms to clinics—while promoting responsible and impactful innovation.
AI News + Events @ Cornell
Event Spotlight
AI4AI Seminar Series
The AI4AI seminar series will occur on the second Friday of each month this fall.
The AI4AI Seminar Series is part of the Assured Integrity for AI-Based Software (AI4AI) initiative led by Cornell faculty by Alexandra Silva and Vitaly Shmatikov. The series brings together researchers and practitioners to discuss emerging challenges and advances in AI security, assurance, and trustworthy AI systems.
Location: Computing and Information Science Building room 350, Ithaca Campus
Click here to attend via Zoom
The AI4AI seminar series will occur on the second Friday of each month this fall.
Date: September 11, 2026
Speaker: Guy Amir, postdoctoral researcher, Department of Computer Science, The University of Texas at Austin
Title: Verifiable Fault Tolerance in AI Training
Date: October 9, 2026
Speaker: TBD
Title: TBD
Date: November 13, 2026
Speaker: TBD
Title: TBD
Date: December 11, 2026
Speaker: TBD
Title: TBD
Weak AI regulation may backfire, making products less safe
A new modeling study finds no regulation of AI products and services may be safer than weak regulation.
Can AI plan for heat emergencies better than simple rules? It depends
For consequential decision-making, the benefits of a simple index score vs. a less-interpretable predictive AI algorithm depend, researchers from Cornell found, on the desired outcome as well as the decision’s intended audience.
Amazon partnership establishes Cornell AI security initiative
Cornell computer scientists will lead the development of safety protocols to shore up AI agents and the code they produce.
Connect
We welcome other corporate engagements in AI4AI.
Please contact Laura Batten, Director of Strategic Partnerships, for more information.