Alice raises $140M to secure AI models as revenue nears $100M ARR

Alice has raised $140 million to tackle a problem that gets harder as AI systems gain more autonomy: keeping models and agents from doing things their developers never intended.

The New York and Tel Aviv-based AI security startup announced the funding Tuesday, saying Apax Digital Funds led the round, joined by Samsung, SentinelOne, Maj Invest, MoreTech and Phoenix Insurance. Existing backers Resolute Ventures, Grove Ventures, CRV, Highland Europe, Vintage Investments, Norwest, NFX and Claltech participated. The financing brings Alice’s total funding to $280 million, and Apax Digital will join its board.

The round includes another standout number. Alice says it is approaching $100 million in annual recurring revenue, with its AI business growing more than 500% over the past two years.

Formerly known as ActiveFence, Alice spent years tracking fraud, extremism, coordinated manipulation and other adversarial activity across major internet platforms. Now it is applying that experience to AI, where jailbreaks, prompt injection and autonomous agents have created a new security problem for companies deploying models in production.

“There are infinite ways to break an AI, and you can’t defend against something you’ve never seen,” said Noam Schwartz, CEO and co-founder of Alice. “We spent nearly a decade learning exactly how people abuse technology at the scale of billions, first on the world’s largest platforms and now on the models those same people are probing. We are very quickly democratizing capabilities before we’ve democratized the defenses. This round is about closing that gap.”

AI agents are creating a bigger security target

Alice’s pitch becomes more relevant as companies move AI beyond chatbots and give agents access to databases, internal software, APIs and tools that can take actions without constant human approval. A model producing a bad answer is one problem. An agent with permission to execute actions can create a very different class of risk.

Alice works with frontier AI labs before models reach users, testing them with malicious prompts and agentic tasks intended to expose unexpected behavior. The company named Anthropic, Google and Cohere among its customers and says it works with eight of the 10 leading AI model labs.

For enterprise deployments, Alice lets companies define their own policies, simulate attacks and test for data leaks, compliance failures and unintended behavior. Its systems then monitor model inputs and outputs once those applications are running.

Behind that work is Rabbit Hole, Alice’s proprietary dataset of adversarial and harmful content collected from years of tracking abuse online. Alice says the same underlying technology safeguards more than 3 billion people across platforms that include Google, Meta, TikTok and Amazon.

That historical attack data may be Alice’s biggest differentiator. AI security increasingly resembles an arms race in which defenders must anticipate attacks that have not yet reached a particular model. Alice is betting that abuse patterns observed elsewhere on the internet can provide an early warning.

“Just as the cloud platform shift created a new category of security, the AI platform shift is opening a rapidly growing attack surface that will only widen as enterprises roll out agents. As adversaries increasingly use AI, defenders can turn to Alice for model testing and guardrails built on the largest proprietary data asset of adversarial techniques. We are excited to partner with Noam and the team as Alice defines the category,” said Patrick Kane, Partner at Apax Digital.

Alice plans to use the new capital to develop its AI security platform, grow the research team behind Rabbit Hole and build out its commercial operations. The company already employs more than 150 AI security researchers.

The timing is hard to miss. AI companies spent the past few years racing to make models smarter. As those models graduate into agents capable of acting on behalf of people and businesses, the next race may be making sure they stay within the lines.