Airis Labs exits stealth with $60M to transform video intelligence for defense and government

Airis Labs comes out of stealth with $60M and a video-intelligence pitch to defence agencies

The Tel Aviv-founded defence-AI firm has closed a $31M Series B led by PSG Equity, bringing total funding to $60M as it scales US operations from Washington DC. Airis Labs, the defence-AI startup that has spent the past two and a half years operating in stealth, emerged publicly on Tuesday with $60m in total funding, including a $31m Series B led by US growth-equity firm PSG Equity. The round brings in TLV Partners, Stepstone Group, Redseed Ventures, and angel investors including Eyal Waldman, the former Mellanox co-founder and chief executive whose company was sold to Nvidia for roughly $7bn in 2020. The company's pitch is narrower than the broad "AI for defence" banner under which a generation of startups has been raising capital. Airis builds what it calls a video-first intelligence platform: software that ingests fragmented visual data, security camera footage, drone feeds, body-camera recordings, smartphone uploads, social-media imagery, and the long tail of what it labels "user-generated field intelligence", and produces machine-readable structured output that analysts and AI agents can query, reason over, and act on. The category, on the company's framing, is distinct from traditional video analytics, open-source intelligence platforms, and generic data-fusion tools. The customer base, on its own count, is government organisations worldwide, with named selection into the Oracle Defense Ecosystem. The underlying problem Airis is addressing is real and well-documented in defence-AI circles. Government investigators and military intelligence teams now drown in unstructured visual data. A typical urban-incident investigation can produce thousands of hours of mixed-source footage; a typical drone-mission archive can fill terabytes of unindexed video. Human analysts cannot review that volume fast enough to catch operationally relevant signals before they age out. Building software that turns the raw visual flood into searchable, structured intelligence is the category Airis has bet the company on. The competitive set is meaningful and worth naming. Palantir's Project Maven remains the headline US defence-AI deployment for video analysis. Helsing sits at the European end of the same arc with a broader battlefield-AI scope. Anduril, Scale AI's defence unit, and a clutch of newer Israeli startups including BlueGreen Vision and ION-X compete on overlapping ground. Airis's differentiated position, on its own framing, is that its platform was built in real operational environments from inception rather than productised out of a research lab, with the implication that the user-experience and workflow integration are tighter than peers' demonstrations. The corporate structure is worth pausing on. Airis was founded in 2023 in Tel Aviv by Noam Friedman (chief executive), Rotem Abeles (chief product officer) and Amos Lahav (chief business officer), with a team drawing on alumni from Palantir, Meta, the Israeli Unit 8200 intelligence community and the broader Israeli AI sector. PSG Equity is itself worth a sentence. The firm, founded in 2014, has backed more than 130 software and tech-enabled companies and operates from Boston, Kansas City, London, Madrid, Paris and Tel Aviv. PSG director Rotem Shacham, who led the Airis investment, was previously a principal at Viola Ventures and served in Unit 8200, the same intelligence-corps lineage shared by much of Airis's founding team. The cap-table fit is unsurprisingly tight. What Airis Labs has not yet disclosed is the standard set of substantive questions readers ask of a stealth-exit defence-AI announcement: which specific government agencies are paying customers, what the revenue run-rate looks like, and how the product performs against the named alternatives in operational benchmarks. The company says it is operational in government organisations worldwide and that the platform has been validated by the analysts who use it rather than by benchmarks. That is a thoughtful framing on the qualitative side and a slightly worrying one on the procurement-disclosure side. The next 18 months will indicate which.

AI video analysis startup Airis Labs raises $60M - SiliconANGLE

Airis Labs Ltd., a provider of video analysis software for government agencies, launched today with $60 million in funding. The startup raised just over half the capital through a Series A round led by PSG Equity. The growth equity firm was joined by TLV Partners, Stepstone Group, Redseed Ventures and multiple angel investors. Airis provides a platform that helps public sector organizations such as law enforcement agencies analyze video content. The software can analyze footage from social media, security cameras, drones and other sources. It uses artificial intelligence to combine multiple video fees into a single view of a situation. One of the challenges involved in processing video datasets is that they often lack an index. An index is a file that developers often add to relational databases to speed up queries. It's a collection of shortcuts that reduces the amount of time required to find important information. Without an index, users have to go through an entire dataset to find the specific data points they wish to retrieve. Airis says its platform addresses the challenge. According to the company, the software can speed up some video analysis tasks by a factor of 150. It also removes the need for analysts to use multiple video processing tools. Customers can configure Airis' platform to monitor for potential dangers at a specific location. The software displays events of interest in a feed that shows a natural language description of each incident, where it occurred and when. Users can collect more data about an event using a built-in investigation tool. According to Airis, it provides a drag and drop interface that enables analysts to quickly retrieve specific clips. The underlying AI models can also extract key details from other types of data such as audio snippets and text. Multimodal models, or neural networks that support multiple data types, are often based on the industry-standard transformer architecture. The architecture was originally designed for text analysis, but it can also ingest other files with the help of certain modifications. AI models turn text snippets into mathematical representations called embeddings before processing them. Videos and audio snippets can also be turned into embeddings. As a result, transformers can process such files with the help of an algorithm that turns multimedia data into number sequences. "Government teams do not have a shortage of raw visual data. They have a shortage of machine-readable understanding," said Airis co-founder and Chief Executive Officer Noam Friedman. "The next generation of AI used by government agencies needs to understand the physical world: what happened, where it happened, what changed, what matters and what requires human judgment."