Darwinium, a US-based AI fraud prevention platform, has announced the launch of Agent Intent Intelligence, an intent-based authentication and orchestration solution designed to address security risks arising from the growth of agentic commerce, where AI agents browse, purchase, and manage accounts on behalf of consumers.
The solution is deployed natively at the edge across major CDN providers, including Cloudflare and AWS CloudFront. Rather than blocking or permitting automated traffic outright, it evaluates each request in real time to determine whether it originates from a verified AI agent, a human user, or malicious automation, and applies friction proportionate to the assessed risk.
From bot blocking to intent evaluation
Traditional bot mitigation tools focus primarily on identifying non-human traffic, often defaulting to blanket challenges or outright blocking. As legitimate AI agents become more prevalent across ecommerce and financial services, this approach risks disrupting genuine automated activity while remaining insufficient against sophisticated attacks such as credential stuffing, content scraping, and inventory hoarding.
Darwinium's framework introduces four adaptive response actions, including permitting low-risk agent interactions such as browsing and product discovery, verifying sessions when behavioural signals shift, challenging activity through step-up authentication when financial or account risk emerges, and preventing malicious automation executing high-risk actions. Because enforcement occurs at the edge, controls are applied before fraud propagates downstream, without introducing latency.
Furthermore, the platform supports emerging web standards that allow AI agents to cryptographically prove their identity using HTTP message signatures, which are validated at the edge and combined with behavioural analytics, device intelligence, and journey-level context. The solution also evaluates automation that does not self-identify as agentic, addressing a blind spot in conventional bot defences.
Darwinium has indicated the solution is designed to interoperate with delegated transaction protocols, including Visa Trusted Agent Protocol, Mastercard Agent Pay, and emerging Know Your Agent models.
Alisdair Faulkner, CEO of Darwinium, said the core challenge lies in distinguishing trusted automation from abuse and applying the appropriate level of trust at the right moment.
