Conntour has secured $7 million in new funding led by General Catalyst with participation from Y Combinator to build an AI-powered search engine for security video systems, aiming to make querying camera footage as simple as typing a sentence. The company says the round closed in 72 hours, a brisk pace that reflects surging demand for enterprise-grade video intelligence despite mounting debates over surveillance and privacy.
What Conntour is building for AI security video search
Conntour’s platform lets security teams ask natural-language questions of their camera networks and get precise answers with corresponding clips in real time. Instead of hard-coding rules like “detect motion at Door 3 after 10 p.m.,” an operator can query, “Show anyone in a red jacket leaving a backpack near the lobby desk and walking toward the parking lot,” and the system searches live feeds and archives to surface relevant video segments and generate incident summaries.
The company says its approach fuses vision-language models with a routing layer that selects the most efficient set of detectors per query to minimize compute. That efficiency is core to its pitch: Conntour claims a single consumer-grade GPU such as Nvidia’s RTX 4090 can monitor up to 50 feeds, enabling deployments to scale to thousands of cameras without ballooning hardware costs. The platform can run fully on-premises, in the cloud, or in a hybrid configuration and can integrate with existing video management systems or operate as a standalone stack.
To account for inconsistent video quality—a perennial issue in the field—Conntour returns results with confidence scores so operators can quickly gauge reliability when footage is grainy, poorly lit, or obstructed. The company also supports rule-based alerts for known risks while keeping the conversational layer available for exploratory and investigative use.
A market hungry for searchable video insights
The timing is favorable. Omdia projects that global video surveillance equipment revenue will surpass $33 billion by the middle of this decade as organizations expand camera coverage in logistics, retail, transportation, and public safety. Yet most of that footage is still effectively “dark data,” expensive to store and slow to parse. Comparitech has estimated tens of millions of cameras are deployed in the United States alone, with major campuses and logistics hubs operating thousands of feeds—far beyond what human operators can monitor continuously.
That gap has turned search into the killer feature for security operations centers. By shifting from static watchlists and motion triggers to question-and-answer workflows, teams can cut investigation time and raise detection coverage without hiring swarms of analysts. In practice, this means faster triage of lost-and-found incidents, better reconstruction of complex events across sites, and the ability to retroactively test new risk scenarios against months of recorded footage.
Conntour’s emphasis on compute efficiency also addresses a real budget constraint. Video AI workloads are notoriously GPU-hungry; routing queries to the lightest viable models can reduce total cost of ownership and make edge deployments feasible where bandwidth and power are limited. That design choice matters in warehouses, stadiums, and transit systems where adding a single rack of GPUs per site is a nonstarter.
Ethics and guardrails in focus for AI video surveillance
Public scrutiny of surveillance technology is intensifying, with recent controversies around the use of automated license plate readers and neighborhood cameras sparking fresh concerns from civil liberties advocates. Conntour’s leadership says the company is selective about who it sells to—a stance made easier by early traction with large government and enterprise customers, including Singapore’s Central Narcotics Bureau.
Precision and transparency features can help, but they do not resolve policy questions on their own. Industry organizations such as the Security Industry Association and technical benchmarks from NIST’s evaluations have pushed vendors toward measurable performance and rigorous testing. Confidence scoring, auditability, and strict access controls are increasingly viewed as table stakes for deployments where misuse risks are nontrivial and regulatory oversight is growing.
Funding details and the road ahead for Conntour
The $7 million raise, led by General Catalyst with Y Combinator involved, will fund product development and go-to-market as Conntour pursues larger, multi-site rollouts. The company says a key technical challenge is delivering full LLM-style language flexibility while preserving the efficiency needed to handle thousands of concurrent feeds—an optimization problem at the heart of multimodal AI for video.
With enterprises seeking faster investigations, tighter operational safety, and better incident reporting, AI-native video search is moving from experiment to requirement. Conntour’s bet is that natural-language interfaces, tuned for scale and cost discipline, will become the standard way security teams interact with their camera infrastructure. The funding gives it runway to prove that out—while the industry and its watchdogs continue to debate where, how, and by whom such tools should be used.
