Ai Engineering 3 min read

Cloudflare Client-Side Security Now Open to All Users

Cloudflare expands its Client-Side Security suite to Pro and Business plans, using a cascading AI model to detect malicious scripts and supply chain attacks.

On March 30, 2026, Cloudflare expanded its Client-Side Security suite to all self-serve customers. The update introduces a cascading AI detection architecture to identify malicious browser-side scripts. For developers managing e-commerce sites or handling sensitive user data, this shifts the baseline for defending against modern skimmers and frontend vulnerabilities.

Access and Compliance Monitoring

Cloudflare Client-Side Security Advanced is now directly available to Pro and Business plan users through the dashboard. This removes the previous enterprise sales requirement for advanced features. Users on the free tier receive complimentary domain-based threat intelligence to block connections to known malicious domains.

The Advanced tier includes automated code-change monitoring. If you process payments, this fulfills the specific monitoring mandates of PCI DSS v4 Requirement 11.6.1. It provides immediate visibility into unauthorized JavaScript modifications without requiring third-party auditing software.

Cascading Detection Pipeline

The system processes approximately 3.5 billion scripts daily using a two-stage detection pipeline. High-traffic environments generate significant noise from modern web frameworks. To maintain catch rates without overwhelming security teams with alerts, Cloudflare splits the analysis.

The first stage uses a Graph Neural Network (GNN). The GNN converts JavaScript code into an Abstract Syntax Tree (AST) to perform structural analysis. It evaluates logical patterns to classify the script’s intent. This allows the system to identify data exfiltration routines even within heavily obfuscated code.

Scripts flagged by the GNN trigger a secondary evaluation. Cloudflare passes these suspicious files to gpt-oss-120b, an open-source large language model, for a contextual second opinion. This AI inference runs directly at the edge using Cloudflare Workers AI.

This cascading approach reduces false positives by up to 200x compared to relying on the GNN alone. It preserves the system’s ability to catch zero-day exploits while drastically reducing alert fatigue. Flagged scripts are automatically logged to Cloudflare R2 to support posterior auditing and continuous model improvement.

Target Vectors and Supply Chain Vulnerabilities

Enterprise zones currently run an average of 2,200 external scripts. Small and medium-sized webshops rely heavily on third-party libraries, creating massive attack surfaces. Recent supply chain attacks demonstrate the fragility of modern web development ecosystems.

In September 2025, attackers compromised 18 popular npm packages with crypto-stealing code. By January 2026, researchers found browser-side keyloggers harvesting credentials on a major U.S. bank’s merchandise store. Platforms like Magento are frequent targets for these infections. Without dedicated client-side monitoring, malicious code can operate undetected for weeks.

Review your third-party script dependencies and enable the free domain-based threat intelligence if your domains route through Cloudflare. If your application handles payment details or sensitive user profiles, map your current client-side monitoring against the PCI DSS v4 requirements to determine if the Advanced tier is necessary for your compliance posture.

Get Insanely Good at AI

Get Insanely Good at AI

The book for developers who want to understand how AI actually works. LLMs, prompt engineering, RAG, AI agents, and production systems.

Keep Reading

Ai Engineering

How to Use Symbolic Execution for Automated BPF Analysis

Learn how Cloudflare uses the Z3 theorem prover to instantly generate magic packets and reverse-engineer BPF bytecode for security research.

Ai Engineering

Pangram 4 Replaces Watermarks With Style Analysis in $9M Round

Pangram secured $9 million to launch AI detection models that identify machine-generated text and images without relying on metadata or watermarks.

Ai Engineering

Gemini 4 Carbon reaches the API: $2/$10 until November 30, then double, and a safety report that calls it no material change

Google opened Gemini 4 Carbon to paying developers and AI Ultra subscribers on November 10, with a 1M-token window and a price rise to $4/$20 on November 30. Its safety report carries Argon's risk findings forward, while showing a 12.8-point jump in exploit success.

Ai Engineering

Gemini 4 Carbon nine days on: top of the index, a third of Claude's task cost, and still no API

Google's October 28 frontier model leads Artificial Analysis at 59 and costs about $16 per GDPx task against Opus 5.5's $51. Developers still cannot call it, its hallucination rate rose on one test, and its linked safety report reportedly covers Argon.

Ai Engineering

Cloudflare's GLM-5.3 Bug Leaked Other Customers' Prompts Through Stale KV Cache Pages

Cloudflare disclosed October 26 that a KV-cache pool misconfiguration plus an SGLang bug let 155 prompts longer than 681,280 tokens receive responses attending over other customers' cached context between August 27 and September 9, with researcher Gadi Evron reporting it via HackerOne.