In 2026, the business of watching video feeds is both a booming industry and a minefield of ethical questions. Recent controversies, from federal agencies accessing private camera networks to police requesting footage from residential devices, have intensified debates over surveillance. Yet, the market for smarter monitoring tools continues to expand, fueled by advances in AI that can interpret video.
Conntour, a startup building an AI search engine for security systems, is navigating this tension with a selective approach to clients. Co-founder and CEO Matan Goldner says the company's existing contracts with major entities, including a Singaporean narcotics bureau, provide the leverage to choose partnerships they deem ethical and legal. "We're in control of who is using it and for what," Goldner stated in an interview.
This strategy hasn't slowed investment. Conntour just closed a $7 million seed round from General Catalyst and Y Combinator, among others. Goldner said the funding was secured in just 72 hours.
The platform allows security staff to search live or recorded footage using plain language—asking, for instance, to find "someone in sneakers passing a bag in the lobby." It uses vision-language models to identify objects, people, or situations, a shift from older systems reliant on rigid parameters. It can also generate alerts and reports.
A key claim is scalability: the system is engineered to handle thousands of camera feeds efficiently, monitoring up to 50 streams on a single consumer-grade GPU. It works by dynamically selecting the least computationally intensive models for each query. The software can be deployed on-site, in the cloud, or as a hybrid, integrating with existing infrastructure.
Conntour acknowledges a perennial industry issue: poor video quality from bad cameras or lighting. Its response is to attach a confidence score to search results, flagging potential inaccuracies.
Looking ahead, Goldner identifies a core technical challenge: marrying the open-ended flexibility of large language models with the extreme efficiency needed to process vast video arrays in real time. "This contradiction is the biggest technical barrier in our space," he said.
Source: TechCrunch