AI Agents Challenge Web's Trust Model, Demanding New Identity Frameworks
Cloudflare argues that the rise of AI agents acting on behalf of users necessitates a fundamental shift in how websites manage identity, reputation, and accountability.
The internet is rapidly evolving with the emergence of AI agents performing tasks on behalf of human users. These sophisticated automations, distinct from traditional bots, can mimic human behavior and operate at scales previously unimaginable, posing significant challenges to website operators. Cloudflare's recent analysis highlights that these "Agents," often managed by third-party "Operators," can execute any action a human user could, but without human limitations like fatigue or the aversion to friction.
Traditionally, websites rely on three pillars for security and fraud prevention: identity, reputation, and accountability. These allow sites to verify if a user is truthful about their identity, assess their trustworthiness, and determine who is responsible for their actions. However, AI agents act as intermediaries, sitting between the user and the website. This intermediation obscures the direct relationship, making it difficult for sites to distinguish between individual users and the agents acting for them, or even to identify the agent and its operator.
This shift is problematic because much of the web's infrastructure and business models are built on the assumption of human-scale interaction. Websites plan capacity, monetize services, and manage risk based on predictable human behavior. When agents operate at scale, they can overwhelm systems, facilitate fraudulent activities, and create scenarios where accountability is impossible to assign. For instance, a grocer can no longer recognize a regular customer's preferences when an agent places multiple orders, and a bank struggles to verify a transaction dispute when only the agent's activity is visible.
Cloudflare posits that blocking agent traffic is becoming an untenable strategy as agentic traffic is projected to soon eclipse traditional human browsing. Instead, the company advocates for the development of new mechanisms within the internet ecosystem to preserve the essential relationship between users and websites, even when mediated by AI. This requires ensuring that agents and their operators are recognizable and accountable, just as human users are.
The core principle proposed is the "Intermediation Principle": the business relationship between users and sites, and the associated security responsibilities, must be upheld regardless of intermediation. This means websites need new tools to recognize, serve, and hold accountable users acting through agents, similar to how they handle direct human interactions. The goal is to enable an internet where users, whether human or agent-driven, have at least equal, if not greater, access to websites than before, while maintaining robust security and accountability.
While local agents running on a user's device are considered out of scope for this particular discussion, the focus remains on remotely hosted agents. These agents, by their nature, create new risks even for previously benign behaviors due to their scalability. The challenge lies in adapting existing trust models to this new paradigm without hindering the potential benefits of AI-driven automation.
Cloudflare's vision is not to force agents to act like humans, but to equip websites with the means to obtain or require information about the user behind the agent, enabling accountability. This could involve new forms of verification or identification that are compatible with agentic operations, ensuring that the benefits of AI agents can be realized without compromising the integrity and security of online interactions.