June 2026 Summaries

Fastly, a key player in powering global live events, particularly excels during the World Cup, where its infrastructure supports massive viewership and provides insights into global trends. By analyzing traffic through Fastly's platform, the company observes significant patterns such as a 24.65% average dip in bandwidth during halftime, indicating viewer engagement changes. The presence of celebrity players like Cristiano Ronaldo significantly boosts viewership, as seen in Portugal's match against the Democratic Republic of Congo, which outperformed others in traffic peaks. Matches scheduled earlier in the day generally attract more viewers, although exceptions such as Mexico v. South Korea challenge this trend. Competitive matches tend to see increased traffic, particularly in the latter stages, as seen in the Spain v. Cabo Verde match, which experienced a 33% surge in peak bandwidth. Fastly highlights its role in these events and invites further engagement and questions from the audience as the World Cup progresses, emphasizing its commitment to supporting large-scale, reliable, and secure digital experiences.

Modern applications often require traffic routing across multiple services and domains, which traditionally involves complex configurations and custom code. Request Routing streamlines this process by allowing users to define path-based rules within a Domain Management system, enabling them to host various services like APIs, static assets, and media files under a single domain without the need for custom code. It supports use cases such as microservices orchestration, geographic localization, and A/B testing through header-based matching, with planned support for geographic and IP-based conditions. The system operates on a "Draft-to-Deploy" workflow, which includes stages for drafting, configuring, previewing, activating, and rolling back changes, ensuring efficient and safe deployment. Request Routing is accessible through all plans, offering a user-friendly interface and documentation for those looking to implement or explore advanced options.

Digital publishers are facing a significant challenge with the rise of AI systems that can ingest and summarize their content, providing direct answers to users without generating traffic or revenue for the publishers themselves. This new dynamic disrupts the traditional publisher-search engine relationship, where content was indexed to drive user traffic back to the source, thus benefiting both parties. The AI systems bypass this by delivering direct responses, which results in no visits, impressions, or subscriptions for publishers, effectively eroding their audience ownership and increasing infrastructure costs without corresponding revenue. To counteract these challenges, publishers need to develop a new content control architecture that allows them to manage access, protection, monetization, and measurement of their content. This involves distinguishing between trusted and abusive bots, charging commercial AI systems for content use, and employing sophisticated deterrents like deception techniques to mitigate unauthorized access. Effective management of AI interactions, visibility into content usage, and real-time decision-making about access at the infrastructure edge are crucial for publishers to maintain control and derive value from their content in the AI era.

High-performance infrastructure, while vital for seamless operations, often suffers from a lack of visibility due to its silent efficiency, leading to challenges in justifying investments during budget reviews and strategic discussions. Fastly addresses this "silent performance" problem by redesigning its Account Summary page to present key metrics like Origin Offload and Mitigation Rate in business-friendly terms, highlighting infrastructure cost savings and revenue protection. This approach transforms raw technical data into a narrative that resonates with cross-functional stakeholders, allowing engineering leaders to demonstrate the tangible business value of their investments. By providing proactive insights into traffic trends and automated reporting features, Fastly empowers engineers, executives, and procurement teams to make informed decisions based on shared, easily accessible data, ultimately aligning infrastructure performance with business outcomes without the burden of manual data consolidation.

Fastly's capacity planning model centers on the Gini coefficient, a metric traditionally used in economics to measure inequality, which surprisingly proved effective in predicting traffic behavior during major events. This approach emerged after standard AI and ML techniques failed to accurately forecast capacity demands on unpredictable, high-traffic days. The model treats traffic inequality as a critical signal, linking it to cache behavior and CPU utilization to determine headroom for Fastly's Points of Presence (POPs). By focusing on the concentration of workloads, the model optimizes cache efficiency, allowing for more accurate scenario analysis and capacity planning. The simplicity of this model, utilizing robust regression techniques, enables Fastly to address diverse traffic challenges, leading to more efficient infrastructure investment decisions and traffic management strategies.

Integrating C++ at the edge, particularly through Fastly Compute with WebAssembly (Wasm), offers unparalleled performance and security advantages for developers leveraging legacy code, modern gaming logic, and agentic systems. The newly available C++ SDK enables teams to optimize latency-sensitive applications by capitalizing on C++'s extensive library ecosystem and Fastly's memory-safe Wasm execution. This setup allows for ultra-low latency through efficient Wasm bytecode, ensuring high-speed processing for critical applications like gaming features, anti-cheat solutions, and data parsing without sacrificing security or polyglot collaboration. Additionally, Fastly's Object Storage and Image Optimization services enhance the delivery and transformation of game assets and web content, supporting the rich ecosystems of modern gaming. The introduction of C++ at the edge empowers developers to maintain the language's inherent speed while benefiting from the advanced performance and security features offered by Wasm.

Google Cloud Service Extensions, in partnership with Fastly's Next-Gen WAF, presents a powerful solution for real-time traffic inspection directly within the load balancer pipeline, enhancing security for modern cloud architectures. This integration enables organizations, such as an e-commerce platform on Google Kubernetes Engine, to maintain robust protection against threats like credential stuffing and SQL injection while achieving sub-200ms response times. By leveraging Envoy's external processing protocol and gRPC, the Next-Gen WAF can inspect incoming request headers, payload data, and outbound responses to block threats before they reach the application backend. The deployment process involves creating an application load balancer that supports extensions and utilizing a ready-made Docker image for the WAF agent, ensuring zero-latency security decisions and automatic scaling with existing infrastructure. This architecture offers a cloud-native, enterprise-grade security solution suitable for demanding environments, emphasizing the future of web application security integrated with cloud infrastructure.

In the realm of web security, identity often serves as a digital façade rather than an accurate representation, with attackers frequently exploiting User-Agent spoofing to mimic legitimate browsers and bypass security measures. This blog delves into the mechanics of such deception, illustrating how tools like Python, cURL, and SQLMap can disguise automated traffic with realistic browser headers. It highlights how the Fastly Next-Gen Web Application Firewall (WAF) combats these threats by employing SmartParse technology and TLS/JA3 fingerprinting to analyze request behavior and identify discrepancies between claimed and actual client identities. Unlike traditional regex-based filters, Fastly's WAF utilizes near real-time monitoring, behavioral signals, and deep packet inspection to detect and block malicious requests, offering a sophisticated defense that examines the "DNA" of web traffic to uncover deceitful activities. The blog underscores the importance of moving beyond static identification techniques, advocating for advanced behavioral analysis to gain true visibility into network threats, and invites readers to explore the capabilities of Fastly's Next-Gen WAF through a trial or demo.

Gaming companies face significant challenges in balancing the need for robust security against the desire to maintain seamless player experiences, particularly when it comes to authentication workflows. Credential stuffing attacks, which exploit reused passwords from unrelated data breaches, are a persistent threat, targeting valuable gaming accounts that contain virtual currency, rare items, and linked payment methods. These attacks are difficult to detect because they blend with normal traffic using sophisticated automation tools, especially during times of high player activity like game launches or live events. Traditional security measures, such as static rules and CAPTCHA challenges, can create friction for legitimate users, leading to failed logins and disrupted gameplay experiences. To address these challenges, modern mitigation strategies emphasize real-time decisions at the edge, using behavioral analysis, adaptive rate limiting, and bot detection to distinguish between genuine players and automated threats. Fastly offers solutions that allow gaming companies to protect against account takeovers by detecting and mitigating suspicious activities closer to the source, thus preserving player trust and ensuring smooth, responsive gameplay without compromising security.

Firefox 149, released by Mozilla on March 24th, introduced a free built-in VPN that requires no additional software and features a simple toggle to mask users' IP addresses. The underlying infrastructure of this VPN is based on the MASQUE protocol, which uses QUIC to enable efficient and secure proxy services without revealing user activity to intermediaries. This architecture allows multiple independent data streams within a single connection, minimizing disruptions caused by packet loss, and supports both TCP and UDP traffic. Mozilla's choice of Fastly as its proxy partner is based on Fastly's capacity to handle large-scale traffic and its commitment to user privacy, with no commercial incentives to track or share user data. The incorporation of MASQUE in Firefox's VPN is part of a broader initiative to embed privacy features directly into the browser, heralding a shift toward more secure, native internet privacy solutions.

Live streaming large-scale events such as the FIFA World Cup poses significant challenges for engineering and operations teams, as even minor disruptions can lead to substantial social media backlash, subscriber churn, and revenue loss. The complexity of ensuring a seamless live stream goes beyond video delivery, requiring attention to dependencies like entitlement and authorization APIs, digital ad insertion, and cloud infrastructure readiness. Unexpected traffic patterns, particularly regional surges, can strain systems, emphasizing the need for real-time traffic engineering and mitigation strategies. Effective preparation involves simulating real viewer behavior, not just high traffic volumes, and rehearsing failure scenarios rather than relying on hopeful success. Operational readiness includes clear incident command structures, disciplined coordination, and responsive monitoring, with early detection and agility being crucial in managing unexpected issues. Ultimately, successful live streaming is rooted in a comprehensive systems approach that anticipates and mitigates potential points of failure.

As AI agents increasingly represent legitimate users in online transactions, they pose challenges to traditional internet infrastructure and security protocols designed to block automated traffic due to its historical association with malicious activities. Fastly and Skyfire are addressing this shift by enabling businesses to verify and trust AI agents through Skyfire's agentic commerce infrastructure, which provides verified identity and payment capabilities using Know Your Agent (KYA) credentials. These credentials, similar to Know Your Customer frameworks, allow AI agents to autonomously complete transactions while ensuring they are accountable and identifiable. Fastly’s edge infrastructure facilitates the rapid and scalable enforcement of these credentials, reducing latency and allowing businesses to manage access and transactions efficiently. By integrating Skyfire’s identity and payment verification into their platform, Fastly allows businesses to differentiate between legitimate agent traffic and anonymous automation, providing flexibility in how they manage and monetize agent interactions without the need for extensive reconfiguration.

Fastly has reported a significant increase in AI-generated traffic on its network, with AI requests growing approximately 30% from January to May 2026, outpacing human traffic growth by 6.5 times. This surge is largely driven by two categories: AI crawlers, which methodically scan the web for data to train large language models, and AI fetchers, which execute user-directed tasks such as answering questions or checking availability. While AI crawlers maintain a consistent activity level around the clock, fetchers align more closely with human actions, pointing to an evolving internet layer characterized by intelligent, automated agents. As a result, AI traffic requires more frequent interactions with origin infrastructure, with over 51% of AI requests necessitating direct data retrieval from origin servers compared to less than 9% of human requests. This shift demands smarter web infrastructure strategies to efficiently manage cached content, intelligently route requests for fresh data, and protect origin systems from the high frequency of live information requests. The report highlights the need for businesses to adapt to these changes and leverage them for competitive advantage, with further insights and strategies to be discussed in a subsequent analysis by Fastly's CMO.

AI-driven traffic on the internet is increasing at a pace that surpasses human traffic, posing new challenges for businesses in managing this machine-to-machine interaction. Companies must discern between AI interactions that add value and those that pose risks, recognizing that simply blocking AI traffic is insufficient and can lead to missed opportunities or degraded user experiences. Effective management requires a nuanced approach to each request, leveraging the edge as a real-time decision layer to inspect and enforce policies. Businesses need strategies that focus on visibility, context, and precision to understand AI systems’ behaviors, ensuring they can differentiate valuable automation from harmful activities. Fastly offers solutions that help organizations execute these strategies by providing the necessary tools to balance performance, security, and bot management, ultimately aiding in distinguishing between beneficial and unwanted machine traffic. Forward-looking statements in the text highlight potential risks and uncertainties related to the growth and impact of AI traffic, advising stakeholders to remain aware of these factors as they evolve.

Fastly has introduced a beta release of its Compute Python SDK, allowing developers to run Python code at the network's edge, enhancing speed and security for AI projects and other applications. The SDK cross-compiles standard CPython to execute within isolated WebAssembly sandboxes, supporting popular frameworks like Flask with minimal modifications. Fastly's approach includes memory snapshotting to reduce cold-start delays and maintains Python's dynamic nature while ensuring security through isolated request execution. Although the SDK supports pure Python logic and WASI-compatible extensions, it lacks support for native ML computation frameworks and filesystems. Fastly encourages user feedback and collaboration to refine the SDK as it progresses towards general availability.