Fastly launches AI Firewall and runtime controls to secure LLM traffic

Fastly is trying to become part of the control plane for AI applications by moving security and governance into the network path. AI Runtime Control and AI Firewall effectively turn Fastly’s edge into a policy and observability layer that sits between customers and whatever models or agents they are using. That is strategically important because it decouples AI risk management from any single model provider.

As organizations shift from experiments to production AI, they need to manage token budgets, prompt-injection attacks, model failover and agent behaviour across a messy mix of vendors. If Fastly can make those controls declarative and real time, it reduces one of the frictions that currently slows serious adoption of agentic systems, especially in security-conscious industries.

From a race-to-AGI perspective, this kind of infrastructure quietly advances the timeline by making it safer and cheaper to plug more workflows into increasingly capable models. It also creates a new chokepoint: if edge platforms become the de facto regulators of AI behaviour in production, their policies and interfaces will shape which models win, how quickly multi-model strategies spread, and what kinds of autonomous actions enterprises are comfortable allowing AI agents to take.