Alice Raises $140M to Expand AI Model Testing, Defense, and Monitoring
The AMW Read
A $140M financing materially expands a specialist provider of adversarial model-security data and production monitoring, with implications across AI deployment but no stated market-wide incident or policy shift.
Alice Raises $140M to Expand AI Model Testing, Defense, and Monitoring
Alice, formerly ActiveFence, has raised $140 million in a financing round led by Apax Digital. Samsung and SentinelOne participated. The Israeli company sells AI trust, safety, and security capabilities for model training and production. Its Rabbit Hole dataset contains real-world adversarial and harmful content, which customers use to identify jailbreaks, prompt injections, and related threats to AI systems.
The financing is notable because it directs substantial growth capital to the layer around models rather than to a new model itself. As organizations move models from training into production, testing and monitoring must address adversarial behavior that can change a system's output or steer it toward unsafe actions. Alice is positioning its dataset and security tooling as a persistent control surface across that lifecycle, a role that becomes more relevant as deployment expands from pilots to customer-facing systems.
For builders, the practical implication is to treat adversarial testing and runtime monitoring as product requirements, not a final compliance check; the threats cited by Alice span both model inputs and deployed behavior. For investors, the round is a signal to distinguish security vendors with proprietary threat data and production workflows from generic AI assurance claims. The test will be whether customers turn those capabilities into durable, recurring spend as their model estates scale.
