Milestone Systems announced a suite of artificial intelligence tools for security operations: AI Search, XProtect Video Summarization, and Video Anonymization.
The software is intended to allow security teams to review footage, document incidents, and edit personal data within videos. Milestone framed this announcement to respond to operator concerns regarding accuracy, false positives, and the use of generative AI in environments where decisions can have significant consequences.
Security teams have been slower to adopt generative AI than some sectors, as many still rely on manual review of recordings, written reports, and manual anonymization of footage. Milestone also noted increasing pressure from privacy and AI regulations, as organizations with large amounts of video face questions about how to derive value from those archives within regulations.
Product development
AI Search is designed to help investigators query videos using natural language instead of filters or manual scrubbing. Milestone said operators can describe a vehicle involved in a traffic accident and retrieve matching footage from multiple cameras in seconds.
XProtect Video Summarization generates a text description of your footage that can be used to create incident reports. This tool can run automatically in response to alerts, schedules, or requests by XProtect’s rules engine.
Video anonymization redacts faces and license plates before footage is exported or shared. It is intended for organizations that need to provide video to external authorities or move video between departments without exposing personal information.
Video anonymization will be available in Arcules and XProtect. AI Search is expected to be available by the end of 2026, while an on-premises version for secure, air-gapped environments will be offered through BriefCam.
Data focused
The new tool is built on a vision language model trained on security-specific videos, rather than general-purpose models applied from other fields. Milestone said the model was initially fine-tuned based on use case-specific datasets drawn from transportation scenarios in Europe and the United States.
The company says the underlying video library is collected through fair exchange agreements with customers, anonymized using better AI technology, and tagged with over 300,000 categories using NVIDIA technology. The resulting dataset is traceable and structured to protect privacy, the company said.
Milestone said this data will be used to fine-tune the NVIDIA Cosmos Reason underlying model. In our own tests, we reported a 19.4% increase in flow and direction accuracy and a 4.4% increase in alert validation accuracy compared to the underlying foundation model for our traffic-focused vision language model.
The city of Genoa is the latest contributor to the library, adding video data aimed at increasing the geographic spread of the dataset. According to Milestone, this data foundation is designed to comply with EU AI law and GDPR, two regulations that shape how organizations use video and AI in Europe.
reliable questions
Andrew Barnett outlined the rationale behind the launch in a statement accompanying the announcement.
“How can we build on decades of experience in video management to prepare our customers for the next generation of artificial intelligence? The answer is trust. Trust that the data is compliant, trust that the AI is accurate, trust that these tools truly make operators’ jobs easier. That’s what we’re delivering today,” said Andrew Barnett, Chief Technology Officer, Milestone Systems.
Milestone has more than 1,500 employees and sells video technology used in industries such as manufacturing, airports, law enforcement, retail, and traffic management. The company’s portfolio includes XProtect video management software, BriefCam analytics, Arcules cloud video services, as well as efforts in anonymized video datasets for AI training.
This announcement puts the company more directly in the market where suppliers are looking to apply generative AI to their monitoring and operational software without increasing regulatory risk. Milestone’s pitch is based on two central issues for buyers. One is whether the model can produce reliable output in real-world security workflows, and the other is whether the data used to train and operate the model can withstand scrutiny under privacy laws.
Milestone said its video training library is responsibly sourced, fully anonymized and traceable, and better AI technology is used in internal data pipelines and customer-facing anonymization products.
