เกิดอะไรขึ้น
Ynetnews reports that Alice, formerly known as ActiveFence, raised $140 million in a funding round led by Apax Digital, with participation from Samsung, SentinelOne and other investors. The New York- and Tel Aviv-based company says it provides testing, monitoring and guardrail tools for AI models and autonomous agents.
Ynetnews reports that Alice raised $140 million in a new funding round led by Apax Digital. The outlet says Samsung, SentinelOne, Maj Invest, MoreTech, Phoenix Insurance, Norwest, CRV, Vintage, Grove and Highland Europe also participated. Alice was founded in 2018, is headquartered in New York and Tel Aviv, and was formerly known as ActiveFence. The source describes the transaction as a company announcement made on Tuesday. This article does not independently confirm the amount, the participants or the terms of the round.
According to Ynetnews, Alice says it is approaching $100 million in annual recurring revenue, protects more than 3 billion users online and works with eight of the world’s 10 leading AI model laboratories. The company’s named customers and partners include Anthropic, Google and Nvidia. Ynetnews also reports that Alice employs more than 150 researchers. These figures and relationships are presented as claims by Alice or statements reported by the outlet; the source does not provide audited financial statements, contracts, customer attestations or independent measurements supporting them.
Ynetnews reports that Alice began by identifying fraud, extremism, coordinated manipulation and other harmful activity across major online platforms, and is now applying that experience to AI security. Its researchers reportedly simulate malicious prompts and tasks during model development and training to look for jailbreaks, and other failures. After deployment, the company says its tools allow organizations to set safety policies, test for data leaks and compliance failures, and use real-time to monitor inputs and outputs. Alice says its Rabbit Hole product is a proprietary database of adversarial and harmful online activity collected from real-world platforms, used to design tests and update defenses. The source does not describe the database’s size, coverage, validation process or independent audit.
รายละเอียดที่มา: ynetnews.com ↗
ทำไมมันถึงสำคัญ
The funding reflects growing demand for security systems designed specifically for AI models and agents that can browse the internet, retain information and take actions. Alice’s reported approach combines adversarial testing before deployment with real-time safeguards after deployment, although the source does not independently verify the company’s customer, revenue or performance claims.
The company’s chief executive and co-founder, Noam Schwartz, told Ynetnews that model builders need to trust systems to perform their intended functions, requiring regular stress testing and defenses informed by real attacks. Ynetnews reports that Alice plans to use the new capital to expand its AI security platform, increase investment in Rabbit Hole and grow its commercial operations serving AI laboratories and enterprises. The source does not say how the funding will be divided, identify specific hiring targets or provide a timetable for product expansion. Ynetnews places the fundraising in the context of a shift from chatbots toward AI agents that can browse online, retain information and independently take actions.
The outlet says Alice cited the 2026 International AI Safety Report and research by the nonprofit METR as evidence that protected models can still be compromised and that agents may act outside assigned tasks. Those references are reported as material Alice used to explain the need for its products; the source does not independently establish the cited findings or connect them to a specific Alice customer incident. AI security is becoming a distinct commercial concern because model failures can involve more than inaccurate answers. A system that can call tools, access data or take actions may create security and compliance problems when it follows malicious instructions, leaks information or operates beyond a user’s intent.
Ynetnews’s account presents Alice as covering testing during development, policy checks before release and monitoring after deployment. That breadth could matter to organizations that otherwise buy separate tools for model evaluation, content safety, application security and operational monitoring. The reported funding also indicates that investors see AI deployment as a potentially large security market. Apax Digital partner Patrick Kane told Ynetnews that the expansion of AI platforms is creating a wider attack surface as enterprises roll out agents. That is an investor’s assessment, not independent evidence of market size or Alice’s competitive position. The practical question is whether companies will pay for continuous testing and protection as AI systems become connected to internal databases, software tools and external services.
กลไกเชิงโต้ตอบ: มันทำงานอย่างไร
สำรวจเทคโนโลยีเบื้องหลังการพัฒนานี้แบบโต้ตอบ
crm_get_transaction(id='4092').Why can ethical evaluation not be reduced to one model score?
จะดูอะไรต่อไป.
Alice plans to expand its platform, invest further in its Rabbit Hole adversarial-activity database and grow its commercial operations. Key unknowns include the company’s independently verified revenue, the identities and scope of its customer relationships, the measurable effectiveness of its defenses and whether its tools can keep pace with new attacks against increasingly autonomous systems.
Alice’s background in online abuse monitoring may give it a source of adversarial examples, but the relevance of those examples to modern AI systems remains an open question. Harmful activity on internet platforms and attacks against models can overlap, yet they are not identical problems. Ynetnews reports that Alice’s Rabbit Hole database feeds observed attacks into new tests and , but it gives no details about how the company distinguishes relevant threats, measures false positives or prevents defenses from becoming outdated. The source’s description of Alice’s customers and partnerships is broad, and it does not specify which products are deployed, in what environments or with what results.
There are no independently reported benchmark scores, incident-prevention statistics, customer testimonials or examples of vulnerabilities found before release. Those omissions do not disprove Alice’s claims, but they limit what can be concluded from this report. In particular, the funding round should not be treated as evidence that Alice has prevented models from going rogue or that its defenses work across all AI systems. Watch for independently verifiable information about new products, public evaluation results, documented customer deployments, hiring and geographic expansion, and clearer descriptions of Rabbit Hole’s data and testing methodology. It will also be useful to see whether Alice publishes evidence about how its tools perform against jailbreaks, , data leakage and agent behavior under realistic conditions.
A second issue is how Alice’s tools fit into the growing AI security ecosystem. Ynetnews reports that the company intends to serve both frontier AI laboratories and enterprises deploying AI products, but those customers may have different requirements. Model developers may prioritize pre-release evaluation and red-team testing, while businesses may need access controls, audit trails, policy enforcement and incident response. Future disclosures could clarify whether Alice is selling one integrated platform or a collection of services aimed at different stages of deployment. Finally, the unresolved question is whether continuous can keep pace with systems that gain more tools, memory and autonomy. Ynetnews reports that Alice’s researchers study manipulation, safeguard circumvention and unintended behavior, but the article does not establish how often tests run, how quickly defenses are updated or what happens when safeguards fail. Public evidence on those points, along with independent scrutiny of the company’s financial and customer claims, will determine how much weight to place on this fundraising as an indicator of practical progress in AI security.