News

Updates, milestones, and announcements from the Clarkson AI Institute

Software at Machine Speed: How AI Accelerates Cybersecurity Risk and What It Foreshadows for Software Development

A talk by Pat Wilbur, Instructor, Clarkson University

How is AI changing cybersecurity? How has it already transformed software engineering? And what new responsibilities do computer scientists, software engineers, and technology organizations face as AI systems become more autonomous?

The talk begins with a brief look back at “classical” zero-day defense — anomaly detection, attack-surface reduction, instrumentation, and layered controls — then argues that AI changes the risk environment by collapsing the cost of trial and error. Drawing on Anthropic’s reported Mythos results and Nicholas Carlini’s “Black Hat LLMs,” it considers how frontier models may find critical vulnerabilities, generate sophisticated multi-stage exploits, support privilege escalation, and enable non-experts to perform security work that once required elite expertise.

From there it broadens to the future of software development. As AI moves from prompt-based assistance toward autonomous agents, code production may no longer be the main bottleneck — instead, governance, testing, security review, attribution, rollback, and recoverability become central. The talk also considers how agentic AI “coworkers” are beginning to reshape tech-company org charts, and closes on AI attribution practices — including emerging Linux kernel guidelines — as one way to target human review and rescan AI-assisted code as models improve.

Edited from a broader in-class discussion of computer and operating-system security, the video is intended as an overview for students, software engineers, computer scientists, and anyone interested in how AI is changing cybersecurity, software development, and the responsibilities of technical professionals.

AI · Cybersecurity · Software Engineering · Computer Science · Zero-Day · AI Safety · Operating Systems · LLM · Agentic AI · Secure Software

Read the transcript

Classically speaking — and it’s funny to use that term, because I usually use “classic” to mean decades ago, but I guess classically means a few years ago in the context of security, and especially AI.

Classically, the best way to defend against a zero-day vulnerability would be, in my opinion, to instrument the system so that you can detect when one occurs. Hopefully you can detect anomalies — maybe based on the resources used on a system: spikes in CPU, spikes in disk activity, spikes in network activity, strange accesses to files.

That’s a detection mechanism, so that you can respond after the fact — and also to have multiple layers of security to mitigate the potential attack vectors. The idea behind multiple layers is that in each layer there’s some probability a zero-day vulnerability could occur — meaning some black hat or hacker discovers, sometime after it was introduced, that there’s a vulnerability in the system — and they start attacking it.

Excerpt from the opening of the talk.

Watch on YouTube →

AI Minor Passes Faculty Senate

The Clarkson AI Minor has passed the Faculty Senate and will soon be available to students across the university. This milestone marks a major step forward in Clarkson’s commitment to integrating AI literacy across disciplines.

The Minor in Applied Machine Learning and Artificial Intelligence is jointly offered by the Computer Engineering and Computer Science departments and is designed to be accessible to students across nearly all majors. It adds practical AI and ML skills without requiring additional resources — all courses are already in the catalog.

Learn About the AI Minor →

Stay Up to Date

Members receive updates directly. Join the institute to stay informed about news, events, and opportunities.

Join the Institute