Newsroom
RSS feedNews
Product updates, research and perspectives on AI security and model release integrity.
ProductVraelis Contour status: a private Kubernetes collector prototypeWhat changed on the status page this week, what has been checked, and what has not.
PerspectiveFive ways a model update goes wrong after approval
ResearchA self-report is evidence, not proof
PerspectiveEvidence unavailable is a resultAll stories
8 stories
October 2026
- Vraelis Contour status: a private Kubernetes collector prototypeWhat changed on the status page this week, what has been checked, and what has not.Product
- Five ways a model update goes wrong after approvalApproval is a decision about a release. Each of these failures happens after it.Perspective
- A self-report is evidence, not proofWhat a runtime's own report can establish about a model release, and where it stops.Research
- Evidence unavailable is a resultWhy a security record should keep the observations that confirm nothing.Perspective
- Release integrity for machines in the fieldRobots and vehicles receive model updates far from the people who approved them. What evidence would help.Research
- A read-only evidence app for Vraelis ContourSix screens over the recorded trace, with every finding from internal automated review addressed.Product
- Vraelis Contour, our first product directionModel release security, from approval to reported runtime state.Product
- Vraelis focuses on AI securityCybersecurity software for AI in defense, critical infrastructure and autonomous systems.Company
Press and media
For interviews, background or questions about a story, write to us through the contact form. Follow new stories by RSS.