Cloudflare launches Adaptive Intelligence to make bot attacks too costly to sustain

Cloudflare has launched Adaptive Intelligence, a new bot-detection engine designed around a different goal: not blocking every determined attacker, but making attacks slow and expensive enough that they are no longer worthwhile.

The approach reflects the scale and speed of modern abuse. Cloudflare analyzes more than one trillion requests per day, while attackers can change tools and tactics continuously. Traditional rule-based defenses often update in discrete releases, creating a stationary target that attackers can probe and eventually bypass.

A moving defense

Adaptive Intelligence adds a continuously retrained machine-learning model to the system behind Cloudflare’s bot score. It combines network-wide signals to estimate the likelihood that each request is automated abuse, incorporating new bypass tools and bot frameworks without waiting for a scheduled model update.

Cloudflare says the engine is intended to reduce the feedback attackers use to improve their tooling. It can identify a bot from a signal without visibly reacting, allowing the attacker to continue relying on a weakness they do not know has been detected. Its statistical scoring also considers multiple signals rather than exposing one fixed rule that can be isolated and defeated.

The company describes two additional capabilities planned for the engine: disposable rules that target a specific attack and are retired or changed at random intervals, and learning from the traffic the system protects. Only the continuously retrained machine-learning component is launching now.

For teams building AI-powered products, the announcement highlights a broader security shift. Static safeguards can become training data for sophisticated automation; defenses that adapt continuously may limit the value of each probing attempt while reducing the need for manual rule updates.

Source: CloudFlare