AI aggregator 302.ai left 17.6 million user files, including chat logs, exposed in an unsecured Azure bucket
Cybernews researchers found that AI aggregator 302.ai, developed by Hong Kong-based Univerads Technology Limited, left 17.6 million files (22.72TB of user data) publicly accessible in an unsecured Azure storage bucket. The data included more than 144,000 chat logs, some containing personal details such as birth dates, contact details and home addresses, plus enterprise client data such as conversations from an unnamed sex doll maker's customer service bot. The leak was found on August 2 and flagged on August 11, but the bucket was not observed closed until September 8. Cybernews said it saw nothing to suggest bad actors had viewed or copied the data.
What the AI did
The AI aggregator 302.ai stored users' prompts, model replies, chat logs, attached files and generated images, video and audio in an Azure storage bucket that was left unsecured and publicly accessible. This included conversations from a business customer's customer service bot.
First reported October 21, 2026 · Added to the register November 1, 2026 · 2 sources
- Developer
- Univerads Technology Limited
- Model
- 302.ai
- Affected
- Unnamed well-known sex doll maker
- When it happened
- Not stated in the sources
- First reported
- October 21, 2026
What this means for you
Could this affect you?
If you or your vendors use 302.ai. Individual and business users of 302.ai had their prompts, outputs, uploaded files and generated media stored by the service, including businesses running customer-facing chatbots through it, and 17.6 million of those files were left open to anyone.
What to check
- Map which AI intermediaries and aggregators store your AI conversation data.
- Include those services in your vendor and data protection due diligence.
- Require vendors to configure storage securely and respond promptly to breaches.
- Avoid sending personal or sensitive data through third-party AI services without contractual safeguards.
Areas of your AI programme this touches
Timeline
- October 21, 2026First reported
- October 22, 2026Cybernews researchers reached out to Univerads Technology but did not hear back; in the meantime the company locked down the database, according to TechRadar.[2]
Every fact and its source (9)
- Research publishedOctober 21“according to research published by Cybernews on October 21”[1]
- Bucket observed closedSeptember 8“the bucket was not observed closed until September 8”[1]
- Leak discovered and flaggedDiscovered August 2, flagged August 11“the leak was discovered on August 2 and flagged to the company on August 11”[1]
- Conversation archive files exposedMore than 26,000 .zip files“The bucket also held more than 26,000 .zip files of conversation data”[1]
- Chat logs exposedMore than 144,000“More sensitive were more than 144,000 chat logs, stored as .json files”[1]
- Personal data in logsBirth dates, gender, contact details, home addresses“ranging from birth dates and gender to contact details and home addresses”[1]
- Evidence of malicious accessNone seen by Cybernews“it had seen nothing to suggest bad actors had viewed or copied it”[1]
- Size of exposed data17.6 million files, 22.72TB“17.6 million files, totalling 22.72TB of user data, publicly accessible through an unsecured Azure storage bucket”[1]
- Company responseUniverads did not reply; database locked down“Cybernews’ researchers reached out to Univerads Technology, but did not hear back.”[2]
Sources
- Millions of AI files exposed in leak that caught a sex doll maker's chatbotinsurancebusinessmag.com · October 21, 2026
- Security breach at AI aggregator exposes over 18 million files, including some very embarrassing chat recordstechradar.com · October 22, 2026
How this record is classified. Severity S (4/5): real harm to one organisation, data exposed, or a containment failure. OECD level: incident, an event in which an AI system led to actual harm. Evidence: Reported, meaning press reporting only.
The facts in At a glance and in Every fact and its source are quoted from the sources listed. The summary, What the AI did and What this means for you are written from those sources and checked against them automatically before publication. Records reflect what has been disclosed, not everything that has happened. Incidents are found through news feeds and the GDELT Project.
