PulseAugur
EN
LIVE 22:08:35

Cloudflare Workers AI leaks customer prompts due to KV cache bug

By PulseAugur Editorial · [1 sources] ·

A bug in Cloudflare's Workers AI service, specifically within the KV cache mechanism for the GLM 5.3 model, inadvertently exposed parts of previous customer prompts to subsequent users. This memory bookkeeping error occurred between August 27 and September 9, 2026, affecting prompts exceeding 681,280 tokens. The issue stemmed from the SGLang serving engine not properly clearing memory pages after requests, leading to data leakage when prompts exceeded the allocated KV cache size. Cloudflare has since updated its systems and provided recommendations for preventing similar incidents, including verifying KV pool sizes and upgrading SGLang. AI

IMPACT Highlights the critical need for robust memory management and prompt isolation in multi-tenant AI inference services to prevent data leakage.

RANK_REASON A bug in a specific AI service led to a data leak, impacting users of that service.

Read on dev.to — LLM tag →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Cloudflare Workers AI leaks customer prompts due to KV cache bug

How we ranked this

Signal score
6 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
A bug in a specific AI service led to a data leak, impacting users of that service.
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
product, safety
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Same-day
Cluster formed today. Ranking reflects the current source set at time of score.

Full methodology in our editorial standards.

COVERAGE [1]

  1. dev.to — LLM tag TIER_1 English(EN) · Abdul Rehman ·

    How a KV cache bug exposed other customers' prompts

    <p>For about two weeks this summer, a small number of people using Cloudflare's GLM-5.3 model on Workers AI got answers stitched partly from someone else's prompt. Nobody stole a password. A memory bookkeeping mistake in the KV cache let 155 responses read pages a previous custom…

RELATED ENTITIES

RELATED TOPICS