Firefly raises $20 million to turn existing cameras into AI safety systems

Firefly, a New York-based start-up that builds artificial intelligence (AI) software for physical safety, came out of stealth on Tuesday, 20 October 2026 and said it has raised $20 million in a Series A funding round. "Coming out of stealth" means a company publicly announces itself and its product after working quietly. The round was led by Definition Capital, with BoxGroup and Susa Ventures taking part.

Most large buildings, such as college campuses, hospitals and offices, already have security cameras, staff radios and alarm sensors. These systems usually work separately. A guard may watch dozens of camera feeds, a staff member may call for help on a radio, and an alarm may ring in another room, with no single place where all of this comes together. Firefly's pitch is to add a software layer on top of this existing equipment rather than ask customers to replace it.

According to the company, Ember pulls information from the connected devices and sensors to build a picture of what is happening, so that it can detect an incident, give staff the right context and then organise the response. In plain words, it is meant to work as a control room assistant: noticing that something is wrong, telling the right people, and helping them act together. These are the company's descriptions of its product; independent figures on how accurately the system detects incidents were not part of the announcement.

The funding also shows where investors are putting money in AI. Much of the attention in 2026 has gone to chatbots and AI agents that work on screens. Firefly belongs to a smaller group often called "physical AI": software that reads signals from the real world, such as video, sound and location data, and helps people respond to events in buildings and public spaces. The company's founders come from this field. According to the company, Buckley previously scaled the alarms business at Verkada, a building security company, and Penn, an MIT graduate, spent nearly a decade at Apple leading engineering teams.

For Indian readers, the model is relevant because Indian campuses, hospitals and offices also run large numbers of cameras and radios that rarely talk to each other. Any similar system used in India would process video and other information about identifiable people, which falls under the Digital Personal Data Protection Act, 2023. Founders building safety technology in India would need to plan for consent, data storage and access rules from the start.

Short-term: Firefly will use the new money to hire engineers, machine-learning specialists and sales staff and to build new products, the company said. Its first public customer, Front Range Community College, uses the system to help safety staff monitor and coordinate responses across three campuses.

Long-term: The company plans to expand from education into healthcare and enterprise customers. If it can show that its software works on top of existing cameras and radios, it may offer buildings a cheaper route to modern safety systems than replacing equipment. Whether it raises a Series B in 12 to 18 months, as one report suggested, will depend on how well it grows in these sectors.

Who is affected: Safety and security teams at colleges, hospitals and large workplaces, who may get one screen for alerts from many devices; students, patients and staff, whose safety and whose personal data are both involved when cameras and sensors are analysed by AI; and other safety technology start-ups, which now face a new funded competitor.

Questions and Answers