Alice Raises $140M to Pair Model Testing With Enterprise AI Guardrails

Alice has raised $140 million to expand an AI security platform built for two different moments: before a model is released and after an enterprise puts it to work. The company’s bet is that adversarial behavior observed elsewhere online can strengthen both model tests and production safeguards.

Apax Digital Funds led the financing, which Alice says brings its total funding to $280 million. MoreTech, Phoenix Financial, Resolute Ventures, Grove Ventures, CRV, Highland Europe, Vintage Investments, Norwest, NFX, and Claltech participated. Apax Digital will join Alice’s board.

Alice, formerly known as ActiveFence, plans to put the money into its AI platform, the team behind its Rabbit Hole intelligence dataset, and go-to-market expansion for model labs and enterprises. The company is headquartered in New York and Tel Aviv.

Before release, after deployment

For foundation-model developers, Alice offers pre-release stress testing. Researchers simulate malicious prompts and agentic tasks to find unexpected behavior and harden models against jailbreaks and prompt injection, where instructions attempt to override a model’s intended rules.

Once a system is live, the company says customers can set internal policies, run attack simulations, and monitor inputs and outputs in real time. That shifts the task from improving a general-purpose model to enforcing the specific compliance, data-handling, and behavior requirements of the organization using it.

A web-abuse archive becomes the proposed advantage

Rabbit Hole is the link between those stages. Alice says it built the proprietary dataset over nearly a decade by tracking fraud, extremism, coordinated manipulation, and other adversarial activity across the open web. Its premise is that earlier attack patterns can help identify hostile behavior against AI systems.

Alice says it has more than 150 researchers studying how AI systems can be manipulated or fail. It also says it works with eight of the 10 leading AI model labs, protects more than 3 billion people online, is approaching $100 million in annual recurring revenue, and has grown its AI business by more than 500% over two years.

Those reach and growth figures are company-supplied. The execution test is whether Alice’s historical attack intelligence produces controls that work across both settings: broad model defenses for labs and locally tailored safeguards for enterprise deployments.