Fastly launches AI controls as machine traffic surges

Fastly has launched AI Runtime Control, AI Firewall and new API Security features for enterprise AI systems, as machine-generated traffic has overtaken human traffic on its network.

The new products target three areas of enterprise AI use: applications calling AI models, users and systems interacting with AI applications, and software agents calling enterprise APIs. They are available on the same platform Fastly uses for content delivery and security services.

AI traffic on Fastly's network grew 6.5 times faster than human traffic over the first five months of the year, according to the company. Machine-generated traffic accounted for more than half of network traffic in July and August, highlighting how quickly AI services and automated systems are consuming internet capacity.

The launch pushes Fastly further into a fast-growing part of the cybersecurity and cloud infrastructure market, where suppliers are trying to address risks linked to the spread of generative AI and autonomous agents. Businesses are under pressure to manage model access, monitor spending and prevent agents from interacting with internal systems outside approved rules.

Traffic controls

AI Runtime Control gives customers a single endpoint for routing model calls across public and self-hosted providers. It also includes virtual keys to protect provider credentials, along with controls for token spending, rate limits, budgets and failover.

The approach is designed to help companies apply central policies across different model vendors without locking themselves into a single provider. Fastly is targeting businesses moving beyond AI trials that need closer oversight of how models are used in live systems.

AI Firewall is designed to stop large language model attacks, including prompt injection, by evaluating prompts in the request path before they reach the model. This allows customers to identify and block malicious requests at the edge rather than after a model has processed them.

The API Security additions focus on software agents and automated tools that interact with enterprise APIs. The service enforces API contracts and can either observe or block requests that do not conform to expected rules, including attempts to use unsupported operations or malformed requests.

Budget pressure

The commercial backdrop is also shaping demand. Fastly cited McKinsey research showing that 93% of organisations are exceeding their AI budgets, suggesting that governance over usage and spend is becoming as important as defending against security threats.

Fastly is positioning the new tools as a way to address both issues at once: controlling model traffic in real time while reducing the risk that agents or users trigger unwanted costs or unsafe actions. That reflects a broader shift in the AI market, where oversight is becoming a central buying criterion as deployments move from prototype to production.

Fastly already sells bot management, DDoS protection, web application firewall products and API discovery tools. The latest additions broaden that portfolio by focusing on AI-specific interactions, from incoming prompts to agent-led calls into business systems.

The announcement also highlights how infrastructure providers are adapting to the rise of machine traffic as a defining feature of internet use. If automated requests continue to outpace human activity, networking and security providers may need to redesign controls around software agents and model calls rather than traditional user sessions.

Fastly operates a global edge network with 622 Tbps of capacity and handles more than five trillion requests a day, according to the company. That scale gives it an existing customer base and delivery platform on which to add AI governance features without requiring customers to deploy separate systems.

Kelly Shortridge, Chief Product Officer at Fastly, said the company sees runtime control as a requirement for wider AI adoption inside enterprises.

"To effectively adopt coding agents and implement AI features without losing velocity or eroding existing business resilience, enterprises need control in production, at runtime, without delay, friction, or disruption," Shortridge said.

"Fastly is extending the real-time control customers already trust for content delivery and software security to the models, applications, and agents shaping this new era of distributed systems. Organisations deserve the freedom to move fast with AI while staying safe, achieving innovation at scale without trading off resilience," Shortridge said.