Harmonic Security, a startup based in San Francisco and London that monitors how employees use AI tools at work, has raised $50M in a Series B round. Crosspoint Capital Partners, a private equity firm that invests in cybersecurity, privacy and infrastructure software, led the round.
Existing investors also took part, though the company did not name them in its announcement. Harmonic said the round was oversubscribed and takes its total funding to $76M. It did not disclose a valuation. The money will go into expanding its engineering, sales and marketing work in the United States and Britain.
Harmonic’s software sits between employees and the AI tools they use, checking what they type or upload before it reaches services such as ChatGPT. It also reports which uses of AI are paying off.
“Every board I talk to is asking two questions: how do we reduce the risk of AI use, and are we getting value?” said Alastair Paterson, its chief executive and co-founder.
What The Software Sees And Stops
The main product is a browser extension, a small add-on installed in the web browser, that works in Chrome, Edge, Firefox, Safari and several other browsers, including AI browsers such as Comet and Dia. Companies can install it on staff computers through device management tools such as Microsoft Intune, Jamf or Kandji.
The extension identifies the account behind each session, so a personal account on an AI service can be treated differently from the company’s approved corporate one. When someone is about to paste sensitive material, it can warn them, ask for a business reason or point them to an approved tool. It can also block the action, although Harmonic argues that blanket bans push staff towards riskier tools.
Harmonic built its own small language models, compact AI models trained for a narrow job, to judge each interaction in context. The company says they do this within milliseconds. In its example, the models can tell a developer referring to a data schema apart from someone pasting live passwords or access keys.
The same models power a feature Harmonic calls Usage Intelligence. It shows managers which uses of AI are creating value and which employees have found faster ways to work. That speaks to the second of the two questions Paterson says boards are asking.
From Browser Tabs To AI Agents
Harmonic started in the browser, but it now also covers desktop AI apps and agentic workflows, where AI software carries out tasks on its own. In October 2025 it launched MCP Gateway, a locally installed tool for the Model Context Protocol (MCP). MCP is an open standard that lets AI models connect to outside apps, tools and company data.
The gateway watches that traffic, shows security teams which AI clients and servers are in use and blocks risky actions. It applies the same sensitive-data checks as the browser product, aiming to stop confidential material and intellectual property leaving the business through AI agents.
Harmonic is also plugging into other vendors’ tools. It says it is among the integrations for Agent Builder, a new feature in Cisco’s Cloud Control platform, connecting through MCP. It has also connected to the Compliance API that Anthropic offers for its Claude AI models. That application programming interface (API) gives security teams activity logs from Claude Platform and, for Claude Enterprise customers, the content of conversations.
Harmonic says its software now covers millions of employees at more than 100 companies, including some of the largest enterprises in the world. It has not named those customers. It has also signed new distribution deals with Climb and Distology, two technology distributors.
Founders Who Sold Digital Shadows
Paterson and Bryan Woolgar-O’Neil, the chief technology officer, founded Harmonic in 2023. Both came from Digital Shadows, a threat intelligence company that Paterson co-founded in 2011 and led as chief executive, with Woolgar-O’Neil as its chief technology officer. The cybersecurity company ReliaQuest bought Digital Shadows for $160M in July 2022.
Harmonic began with a $7M seed round led by Ten Eleven Ventures, a venture firm that backs cybersecurity companies, in October 2023. The venture capital firm Next47 led a $17.5M Series A a year later. At that point the company had 30 staff and said its enterprise customers were in double figures. It now employs more than 90 people across the United States and Britain.
Data Slipping Into Chatbots
Gartner, the research and advisory firm, surveyed 302 cybersecurity leaders between March and May 2025. It found that 69% of organisations suspect or have evidence that employees are using banned public generative AI tools. Gartner predicts that by 2030 more than 40% of enterprises will suffer security or compliance incidents tied to unauthorised “shadow AI”, meaning AI tools staff use without approval.
Harmonic also publishes its own research, drawn from anonymised data collected by its monitoring product at companies in the United States and Britain. Its study of more than three million prompts and file uploads across 300 AI tools, covering July to September 2025, found that 26.4% of file uploads contained sensitive data, up from 22% in the previous quarter.
Harmonic describes the companies in that sample as security-conscious enterprises that use its monitoring. Gartner now tracks this area as a market it calls AI usage control, and Harmonic says it is included as an example vendor in Gartner’s overview of that market.
A Crowded Field Of AI Guards
The cybersecurity company SentinelOne agreed in August 2025 to buy Prompt Security, whose software gives companies control over how staff use AI tools and guards against data leakage. A month later, the network security company Cato Networks bought Aim Security, which offered tools for securing employee use of public AI apps.
Noma Security, which focuses on securing AI agents, raised a $100M Series B led by Evolution Equity Partners in 2025, and Zenity raised a $125M Series C led by Norwest in August 2026 for similar work. Rein Security, from New York and Tel Aviv, Israel, raised $25M for software that aims to stop AI agents acting out of line inside the applications they run in.
Against that field, Harmonic presents itself as a tool that tells security teams both what is going wrong with AI at work and what is going right. Its new backer frames the problem in similar terms.
“Security has too often slowed the adoption of new technology,” said Zach Sivertson, a managing director at Crosspoint. “As organizations embrace AI, they shouldn’t have to choose between moving quickly and maintaining control.”
