Milestone unveils its Vision Language Model, hackathon winner, and 2026 goals

At the Milestone Developer Summit 2025, held at Copenhagen’s historic Carlsberg Museum, video technology took center stage as Milestone Systems showed how its Vision Language Model (VLM) is aiming to redefine how cities, organizations and developers harness video data.

The VLM trained on more than 75,000 hours of ethically-sourced footage, according to a release, and was powered by Nvidia’s Cosmos-Reason architecture. The VLM is engineered to interpret complex urban environments through video. It understands city-specific visuals, symbols, languages, lighting and weather conditions.

The summit gathered developers, technologists and partners from 15 countries, with 55 AI developers having early access to the VLM, culminating in the Project Hafnia Hackathon. Participants were challenged to build third-party integrations using the VLM API to unlock new capabilities in smart city video analytics.

Sebastian Döllner, VP of Technology Partnerships & Open Platform at Milestone Systems, framed the initiative as progressing open video ecosystems: “We’re building more than video management software — we’re building an ecosystem that allows partners and customers to innovate on top of it.”

Milestone also previewed a generative AI plugin for its XProtect Video Management Software. This plugin automatically converts live footage into written reports, summaries and real-time alerts. It is targeting improvements in traffic management across ports, airports and urban areas, the company says, shifting from video storage to video intelligence.

The hackathon showcased the VLM’s versatility. The winning project, “Ask The City” by German developer Thomas Kreutz, transformed live city camera feeds into privacy-aware, real-time answers to natural language questions. Judges from Milestone and Nvidia selected the top three finalists from a shortlist of six.

“I wanted to build something that is relevant and useful in real life,” Kreutz said. “The API and documentation made it simple to build a demo quickly, brainstorm more ideas, and try them out.” His innovation earned him €5,000 ($5,786) and the Nvidia Jetson AGX Thor Developer Kit.

Other finalists demonstrated how video can be reimagined: from emergency response systems that triage incidents in near real-time, to drag-and-drop AI workflows for security integrators, to tools that convert short clips into interpretable insights. Each project illustrated how video, when paired with advanced language models, becomes a dynamic source of intelligence rather than a static archive.

Roland Harwood, Community Lead at Hafnia, commented: “The hackathon success is promising for the future use of our platform and data library to train computer vision models on compliantly sourced, curated, extensively annotated, and anonymized real-world data.”

Milestone Systems targets a video ecosystem and library for AI training

Milestone Systems is strengthening its position in video technology following the acquisitions of Arcules and BriefCam in 2024, alongside the launch of Project Hafnia and the purchase of Brighter AI.

Speaking to International Security Journal (p.13), Milestone’s Jesper Just Jensen, VP of Products, says the expanded portfolio reflects their mission to deliver both on‑premises and cloud‑native solutions whilst furthering their international reach.

Arcules and BriefCam bring cloud‑based video-surveillance-as-a-service (VSaas) and advanced video analytics, respectively, integrated with Milestone’s flagship XProtect video management software. Jensen noted growing demand for cloud‑native services, with the combined offerings enabling scalable, low‑hardware solutions for enterprises.

Project Hafnia is building the world’s largest compliant video library, giving AI developers access to anonymized, ethically sourced footage for training models. The initiative addresses challenges around regulatory compliance, data lineage and access to high-quality video sources.

Brighter AI adds real‑time anonymization technology that removes personally identifiable information (PII). Jensen described it as “a reverse facial recognition process,” allowing organizations to leverage video while meeting privacy and compliance requirements.

Looking ahead to 2026, Milestone plans to make its products easier to deploy across diverse environments, while unifying the strengths of XProtect, BriefCam, Arcules, Project Hafnia, and Brighter AI. The company’s open platform strategy aims to deliver an ecosystem that supports security and broader applications of video technology.

Article Topics

accuracy  |  API  |  biometrics  |  facial recognition  |  Milestone Systems  |  research and development  |  video analytics  |  video surveillance