Alice (formerly ActiveFence)

Alice is the January 2026 rebrand of ActiveFence, an Israeli trust-and-safety company founded in 2018 that spent its first several years protecting social platforms from harmful content — CSAM, extremism, disinformation, fraud — by scanning the open, deep, and dark web for emerging threats. As enterprise and consumer platforms began deploying large language models, ActiveFence found the same adversarial actors turning their attention to jailbreaking and manipulating AI systems, and demand from large model providers pushed the company to build a dedicated AI security line before rebranding entirely around it.

The company’s current product, WonderSuite, covers the AI deployment lifecycle in three parts: WonderBuild tests models pre-launch for jailbreak susceptibility and unsafe outputs, WonderFence enforces runtime guardrails against prompt injection and adversarial inputs, and WonderCheck monitors production traffic for drift and emerging abuse patterns. All three are fed by Rabbit Hole, the company’s adversarial intelligence pipeline that traces back to its original trust-and-safety threat-hunting operation.

Alice says it protects seven of the world’s ten largest AI foundation models and lists Amazon, TikTok, Nvidia, Cohere, and Black Forest Labs among its customers and partners. In January 2026 the company closed a $140 million round led by Apax Digital Funds, bringing total funding to $280 million at a valuation near $1 billion — a large raise for a company still in the early stages of its AI-security pivot.

What sets Alice apart from AI red-teaming startups built from scratch is the years of real adversarial content intelligence it inherited from ActiveFence’s original business, rather than purely synthetic or academic attack generation. That said, the company’s platform and customer roster for the AI-security line specifically are new enough that independent, named case studies of WonderFence or WonderCheck in production are not yet widely available.