Conntour raises $7M from General Catalyst, YC to build an AI search engine for security video systems

The surveillance technology sector is currently facing scrutiny, particularly with recent incidents involving U.S. Immigration and Customs Enforcement utilizing Flock's camera network for monitoring individuals. Additionally, Ring, a home security camera producer, has come under fire for introducing features that allow law enforcement to request footage from homeowners. This controversy has ignited discussions around safety, privacy, and the ethics of surveillance. Despite these challenges, the market for surveillance solutions continues to thrive, propelled by advancements in vision-language models. Matan Goldner, the co-founder and CEO of Conntour, a video surveillance startup, emphasizes the importance of ethical considerations in their business practices. Goldner asserts that Conntour is selective about its clientele, a strategy he believes is feasible due to their established partnerships with significant government and publicly traded entities, including Singapore's Central Narcotics Bureau. "Having substantial customers enables us to maintain control over who uses our technology and for what purposes," Goldner explained in an interview with TechCrunch. "We prioritize moral and legal considerations, carefully choosing customers based on our understanding of how they will utilize our services." This cautious approach has paid off, as evidenced by Conntour's recent funding achievement. The startup successfully raised $7 million in a seed round, which attracted investments from General Catalyst, Y Combinator, SV Angel, and Liquid 2 Ventures. Remarkably, the round was finalized within just 72 hours, reflecting the strong interest from investors. Conntour's innovative platform allows security teams to use natural language queries to search video feeds for specific objects, individuals, or scenarios in real-time, effectively functioning like a search engine tailored for security footage. The system autonomously monitors for potential threats based on predefined criteria and generates alerts as necessary. Unlike traditional surveillance systems that rely on rigid definitions to detect specific activities, Conntour's technology leverages advanced AI models, providing users with enhanced flexibility. For instance, a user can request, "Locate instances of someone wearing sneakers passing a bag in the lobby," and the system will swiftly analyze all recorded and live video to deliver pertinent results. Scalability is a key differentiator for Conntour, with Goldner highlighting that their platform is adept at managing numerous camera feeds simultaneously. Their technology can effectively monitor up to 50 camera feeds using a single consumer GPU, such as Nvidia's RTX 4090, by intelligently selecting the appropriate models and systems to optimize resource use. Conntour's system can be fully deployed on-site, entirely in the cloud, or as a hybrid solution, seamlessly integrating with existing security infrastructures or functioning as a standalone surveillance platform. However, challenges persist in the industry concerning the quality of video footage. Poor lighting or low-resolution cameras can severely hinder the effectiveness of surveillance. To address this, Conntour provides confidence scores alongside search results, indicating the reliability of the footage quality. Looking ahead, Goldner identifies a significant technical challenge: enhancing the capabilities of large language models (LLMs) while maintaining operational efficiency. "We aim to achieve the flexibility of LLM-style natural language processing while minimizing resource consumption," he stated. "Balancing these objectives is the primary technical hurdle we are diligently working to overcome."