Threat Pulse 62High
Cloudflare: GLM-5.3 on Workers AI leaked other customers' prompts in 155 responses
Severity 5ResolvedconfirmedCloud and SaaSAI and software supply chainAI-related

A KV cache misconfiguration plus an SGLang bug let 155 very long requests (Aug 27 to Sep 9) receive pieces of other customers' prompts and generations. Fixed Sep 9, disclosed Oct 26; SGLang fixed in 0.5.20 to 0.5.22. No sign of deliberate exploitation.

· 10 days ago · blog.cloudflare.com
Sourcesblog.cloudflare.com
On X
@DailyDarkWebverifiedOct 27, 14:46 UTC

☁️🚨 CLOUDFLARE: WORKERS AI RESPONSES EXPOSED OTHER CUSTOMERS' DATA Cloudflare says 155 responses from its GLM-5.3 Workers AI model contained data from other customers' requests between August 27 and September 9, 2026. • All affected prompts were longer than

Open on X
@choblin29verifiedOct 27, 16:14 UTC

🚨Cloudflare has disclosed that its AI platform leaked data from one customer's requests into another's responses 155 times over 13 days. The researchers observed invoice numbers, payment dates, personal names, source-code information and AI agent task records

Open on X
Why this score

Plan rubric with caps (verified sweep, Oct 27)

Back to the board

Updated Nov 5, 01:46 UTC. Refreshes on its own.

Add a threat

Paste a link to an advisory, an article or an X post. A person reads every submission before it shows; nothing posts on its own.

Share

The picture people see when the link is posted
Post on XSave picture

This picture is what X, iMessage, Slack and Facebook show under the link. It is drawn fresh from the live board: the board and chart pictures update every 10 minutes, an event picture follows that event.