Alice raises $140M as AI security business grows over 500%

The capital injection highlights a critical shift toward defensive infrastructure as enterprises move from experimental LLMs to autonomous AI agents. For the streaming and media ecosystem, this technology provides a necessary layer of protection against prompt injection and data exfiltration in customer-facing recommendation engines and content moderation tools. As frontier labs like Cohere and Anthropic integrate these guardrails, the industry is moving toward a standardized security stack for generative applications. Watch for whether Alice’s valuation reaches unicorn status in its next round as enterprise agent deployments accelerate through 2027.

Alice enters a rapidly consolidating AI security market where multiple startups and incumbents are vying for enterprise guardrail budgets. In July 2025, SentinelOne announced its acquisition of Prompt Security, an AI security startup focused on real-time prompt injection detection, signaling that endpoint and cloud security vendors are absorbing specialized AI safety tools rather than building them internally. That deal followed a broader pattern: Protect AI raised $60 million in Series B funding in early 2025 to expand its model scanning and supply-chain security platform, which serves enterprises deploying open-source models. These moves suggest Alice's $140 million round positions it among a small group of well-capitalized pure-plays competing against both startups and large security platforms adding AI-native capabilities.

On the business side, Alice's pivot from content moderation under the ActiveFence brand to adversarial AI testing reflects a market shift driven by regulatory pressure. The EU AI Act, which entered into force in August 2024, mandates risk assessments and red-teaming for high-risk AI systems deployed in the European Union, creating compliance-driven demand for adversarial testing services. Alice's Rabbit Hole platform, which automates jailbreak discovery and data-leak simulation, aligns directly with those requirements. Meanwhile, Cohere announced in March 2025 that it had integrated third-party safety layers into its enterprise deployment pipeline, confirming that frontier model providers are increasingly outsourcing guardrail functions to specialized vendors rather than relying solely on in-house alignment teams.

From a technical standpoint, Alice's approach sits within a growing category of automated red-teaming tools that benchmark model resilience at scale. Microsoft released PyRIT (Python Risk Identification Tool) as an open-source framework in February 2024, enabling security teams to programmatically probe LLMs for harmful outputs, establishing a baseline that commercial platforms like Alice must exceed in depth and automation. Independent research from Stanford's Center for Research on Foundation Models found in a 2025 report that fewer than 30% of publicly available frontier models passed adversarial robustness benchmarks without additional guardrail layers, underscoring the addressable market for companies offering external security testing. For streaming platforms integrating generative AI into content recommendation and moderation pipelines, these findings reinforce the need for continuous adversarial evaluation as model complexity grows.