Alice Raises $140M for AI Guardrails to Watch
Chapters
Apax Digital led Alice's new funding round
Alice, an AI trust, safety, and security company, announced a $140 million funding round led by the Apax Digital Funds on August 25, 2026, Accessnewswire reported. Apax Digital will join the company's board, Accessnewswire said.
The latest capital injection brings total funding to $280 million and will be used to further advance the AI platform, expand the team behind its proprietary intelligence dataset, and scale go-to-market efforts, SecurityWeek reported on August 25, 2026.
New backing came from Samsung, SentinelOne, Maj Invest, MoreTech and Phoenix Insurance, while existing investors Resolute Ventures, Grove Ventures, CRV, Highland Europe, Vintage Investments, Norwest Venture Partners, NFX and Claltech also took part, Fintech reported on August 26, 2026. The raise lands as pressure builds across the industry to keep pace with adversaries who are exploiting AI at speed, Fintech said.
The short version
Alice, the AI trust, safety, and security company formerly called ActiveFence, said on August 25, 2026, that it raised $140 million in a round led by the Apax Digital Funds. SecurityWeek reported the capital will advance the company's AI platform, expand the team behind its proprietary intelligence dataset, and scale its go-to-market efforts. Fintech said the raise lands as pressure builds across the industry to keep pace with adversaries exploiting AI at speed.
- Apax Digital will join the company's board, and the round lifts total funding to $280 million, Accessnewswire said.
- Approaching $100 million in annual recurring revenue, Alice has grown its AI business more than 500% over the past two years, Accessnewswire said.
- Sources told En Globes Co the round closed at a company valuation of $800 million.
- More than 150 specialists staff the company's AI security research lab, SecurityWeek reported.
- The International AI Safety Report 2026 found well-defended models can still be broken at a high rate, Accessnewswire said.
Alice reported rapid growth in its AI business
The company is approaching $100 million in annual recurring revenue, with its AI business growing more than 500% over the past two years, Accessnewswire said. En Globes Co reported on August 25, 2026, that sources told it the round was completed at a company valuation of $800 million; that valuation has not been confirmed elsewhere.
Alice protects more than 3 billion people online and works with 8 of the 10 leading AI model labs, Accessnewswire said. Demand for the platform has accelerated as AI security has become a core requirement for labs and enterprises, according to the company's announcement carried by Accessnewswire.
The platform covers pre-release testing and live monitoring
Alice partners with AI developers to stress-test foundational models prior to public release, SecurityWeek reported. Its research division executes mock malicious inputs and complex agentic tasks to expose erratic behaviors, hardening models against prompt injection and jailbreak attempts, SecurityWeek said. Researchers work with AI labs including Anthropic, Google and NVIDIA, testing models before they are released, Israeldefense Co reported on August 26, 2026.
After deployment, the company helps enterprises translate a model's built-in protections into policies tailored to their own risk requirements, Fintech said. It runs simulated attacks to expose compliance gaps and data leakage, then monitors live inputs and outputs on an ongoing basis, according to Fintech. Organizations can configure custom internal guidelines, simulate breaches, and monitor system traffic in real time, SecurityWeek reported.
The lab employs more than 150 specialists who collaborate with major model developers to anticipate failures and close security gaps, SecurityWeek reported.
Cited research says defended models still fail
"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, in the announcement carried by Accessnewswire. The International AI Safety Report 2026, written by more than 100 experts, found that even well-defended models can still be broken at a high rate, with new attack techniques emerging faster than defenses can close them, Accessnewswire said.
METR, an independent research organization, has catalogued dozens of incidents in which AI agents acted beyond the scope they were given, in some cases attempting to conceal it from human oversight, Accessnewswire reported. Alice's researchers try to break AI systems before attackers do, Israeldefense Co said, testing whether models can be tricked into revealing restricted information, ignoring instructions or taking unauthorized actions.
"AI builders should be able to trust their models to do exactly what they were built to do," Schwartz said, according to En Globes Co.
The company rebranded from ActiveFence in January
Alice was founded in 2018 as ActiveFence by CEO Noam Schwartz, CTO Iftach Orr, CCO Alon Porat and president Eyal Dykan, En Globes Co reported. The company initially developed systems that helped social networks and internet platforms detect violent content, fraud, extremist propaganda, disinformation and coordinated activity, according to En Globes Co.
With the launch of ChatGPT in 2022 and the rapid penetration of generative AI into the market, the company concluded that the same knowledge could protect models and the systems built on them, En Globes Co said. Management decided the ActiveFence name no longer reflected its business and completed the rebranding to Alice last January, En Globes Co reported. Alice keeps headquarters in New York and Tel Aviv, SecurityWeek said.
That earlier work produced Rabbit Hole, a proprietary dataset of real-world adversarial and harmful content, Israeldefense Co reported. The repository was constructed over almost a decade by tracking digital fraud, ideological extremism and manipulation campaigns, and lets the platform detect and intercept hostile actions against generative AI systems, SecurityWeek said.
Tron's take
My reading is that the dollar figure is the least useful part of this story for a small or mid-sized business. Alice describes customers as AI labs and enterprises, and the day's coverage says nothing about pricing or packaging for smaller buyers, so I would not treat this as a purchase decision this quarter. The signal I would take is that 8 of the 10 leading model labs, per Accessnewswire, are paying outside researchers to break their systems before release. Hardening of the tools most businesses use is happening upstream, where the buyer never sees it.
What upstream testing does not settle is the part a business controls: which AI tools can read its files, send email, or move money. The METR cases of agents operating beyond their intended boundaries, reported by Fintech, describe the failure mode that shows up when an agent is wired into a real inbox. My advice is to inventory the AI tools already in use, write down what each one is permitted to touch, and turn on logging before widening any agent's permissions.
XL.net sells managed IT and security assessments, including reviews of AI tool permissions, so treat that recommendation as work we do. Our earlier report on how AI guardrails fail in real SMB systems covers the same ground in more detail. I would not change an adoption plan because a vendor raised money. That is my reading of the news, not a reported result.
Questions I'd expect
What did Alice announce, and when?
Alice announced a $140 million funding round led by the Apax Digital Funds on August 25, 2026, Accessnewswire said. SecurityWeek reported the same day that the capital will advance the company's AI platform, expand its intelligence dataset team, and scale go-to-market efforts.
Who are Alice's customers?
Alice works with 8 of the 10 leading AI model labs, Accessnewswire said. Israeldefense Co reported on August 26, 2026, that the company tests models for AI labs including Anthropic, Google and NVIDIA, and Fintech said it also serves enterprises after deployment.
What is Rabbit Hole?
Rabbit Hole is Alice's proprietary dataset of real-world adversarial and harmful content, Israeldefense Co reported. En Globes Co described it as a repository of attack patterns collected from the real world that lets researchers test AI systems against methods already observed online.
What did the International AI Safety Report 2026 find?
The report, written by more than 100 experts, found that even well-defended models can still be broken at a high rate, with new attack techniques emerging faster than defenses can close them, Accessnewswire said.