Fastly, Inc., a US-based cloud computing and edge platform company, has launched three new AI-native capabilities—AI Runtime Control, AI Firewall, and enhanced API Security—delivering live visibility and command over every layer of an organization’s AI activity. Built on Fastly’s unified edge cloud platform, these additions expand the Fastly for AI portfolio to help businesses safely govern usage, block threats, and manage AI agent access as deployments move from trial to mission-critical.
Enterprise AI is scaling at extraordinary speed: machine-generated traffic exceeded 50 percent across Fastly’s network in July and August, while AI traffic grew 6.5 times faster than human traffic from January through May this year. Yet costs are outpacing plans—McKinsey finds 93 percent of organizations exceed their AI budgets—creating an urgent need for real-time oversight of routing, spending, reliability, and permissions. Fastly’s new tools meet this need by applying consistent policies instantly across all model providers and infrastructure, so businesses retain full choice without losing control.
AI Runtime Control centralizes all model traffic through one gateway, tracks token spending live, enforces budgets, manages credentials securely, and shifts traffic automatically to keep services up. AI Firewall blocks prompt injection and other LLM attacks at the edge, stopping threats before they reach models. Updated API Security ensures AI agents follow approved rules when connecting to internal systems, preventing missteps or unauthorized access as new endpoints are built faster than teams can review them.
Fastly CPO Kelly Shortridge said enterprises should innovate boldly without trading off reliability, and these tools extend the trusted real-time control used worldwide for digital delivery directly to AI models, apps, and agents. Running on a global platform with 622 Tbps capacity and handling over five trillion requests daily, this unified framework lets organizations adopt AI at scale safely, securely, and within budget.
