[re]aligned
AI for defense. Show your work.
October 13, 2026 · Online only · A CanSecWest extension
Defenders have been making steady progress with AI. We’re bringing together the people doing that work to show what holds up, explain what fails, and compare notes with the people who have to use it.
Three presentation tracks, plus two Zoom discussion tracks for panels and meetings. Technical talks, working methods, and time to ask the second question.
Included in digital access for all CanSecWest attendees. Online-only passes will also be available after CanSecWest.
The defenders get the floor
The “vulnpocalypse” forecasts have had plenty of airtime. Defenders have been making steady gains too, often through the less theatrical work of testing, verification and putting limits around things.
[re]aligned extends CanSecWest with an online program focused on that work.
We’re interested in detection that gives someone time to act. Code analysis that helps a developer fix the right problem. Agents whose authority can be limited and whose actions can be checked. Incident-response tools that reduce the work without quietly discarding the evidence.
The standard is practical: show the method, explain the comparison, and tell us where it breaks. A good result is welcome. A well-understood failure can be just as useful.
What belongs here
The program will bring together work on:
Detection and threat intelligence: recognising malicious behaviour, testing alerts against attacks, and turning observations into something defenders can use.
Application and supply-chain security: finding and validating flaws, checking generated code, and establishing what evidence supports trust in a dependency.
Agent security: identity, delegated authority, least privilege, isolation, and records that remain useful when something goes wrong.
Incident response and malware analysis: investigation, triage, reverse engineering, and keeping conclusions tied to evidence.
Evaluation and operator practice: realistic tests, false-positive costs, reproducible results, and the skills needed to judge the output.
Work does not need an AI label to belong here. Established defensive methods give us something to test new tools against. We want the people building AI systems in the same discussion as the people who understand the systems being defended.
Talks, with room for a conversation
We’re starting with three presentation tracks and two Zoom panel/meeting tracks.
The presentation tracks give researchers and practitioners room to explain their work in detail. The Zoom sessions let speakers, defenders and attendees question assumptions, compare experience, and work through the parts that do not fit neatly on a slide.
Bring the awkward question about the dataset. Ask what happened when the model was wrong. That is part of the program.
Speaker details and session times will be published as they are confirmed.
Join us on October 13
[re]aligned is online only. There are two ways to attend.
Already attending CanSecWest? Access to [re]aligned is included in your in-person and on-line conference digital access at no additional charge. You do not need a separate online-event pass.
Only joining us online? Standalone passes for [re]aligned will be available after CanSecWest. You will not need a CanSecWest conference pass to attend the October 13 online event. Pricing and the online-only registration link will be posted here when those passes become available.
Come with a problem you are trying to solve, a result worth examining, or a tool you have learned not to trust quite so much. We’ll make room for a useful discussion.