AI2030 Responsible AI Tools Marketplace

Open Source Tools Commercial Tools Research Papers AI Incidents

Chicago Office Address
111 North Wabash Ave.
Chicago, IL 60602
United States

info@ai2030.org

Follow Us

AI2030

Home
About Us
Membership
Sponsorship

Useful Links

Get Tickets
Contact Us
Events

Quick Links

Blog
Speakers
Contacts

Marketplace

Open Source Tools
Commercial Tools
Research Papers
AI Incidents

© 2026 AI 2030 | All Rights Reserved

Privacy Policy|Terms of Use

Back to AI Incidents

Incident #1469·AI Safety & Failures

PocketOS Production Database Was Reportedly Deleted by Cursor AI Agent Running Claude Opus 4.6

April 24, 20263 reportsView on AIID

Explainability

Description

A Cursor AI coding agent reportedly running Anthropic's Claude Opus 4.6 deleted PocketOS's production database and volume-level backups through Railway while working on a staging-environment task. Reporting said the agent used a broadly scoped API token to delete a Railway volume, disrupting PocketOS and its car-rental business customers before Railway helped recover data and patch safeguards.

Actors

Alleged Deployers

pocketos
jer-crane

Alleged Developers

cursor
anysphere
anthropic

Harmed Parties

pocketos-customers
pocketos
car-rental-businesses

Classification

Risk Domain

AI Safety & Failures

Risk Subdomain

7.3. Lack of capability or robustness

MIT Risk Profile

AIUnintentionalPost-deployment

Topics

Explainability

Source

Related Incidents (AI Safety & Failures)

Chatbot Tessa gives unauthorized diet advice to users seeking help for eating disorders

May 2023

Australian Automated Debt Assessment System Issued False Notices to Thousands

Jul 2015

Amazon Alexa Responding to Environmental Inputs

Research Papers on This Topic

Concrete Problems in AI Safety

3,612 citationsJun 2016

A Survey on Explainable Artificial Intelligence (XAI): Towards Medical XAI

2,338 citationsJul 2019

Representation Engineering: A Top-Down Approach to AI Transparency

1,499 citationsOct 2023

Explainable artificial intelligence (XAI) in deep learning-based medical image analysis

1,178 citationsJul 2021

Mitigation Tools

Tools related to this risk domain, via shared responsible AI topics.

Open Source

lime

marcotcr/lime

Lime: Explaining the predictions of any machine learning classifier

⭐ 12,100

facets

PAIR-code/facets

Visualizations for machine learning datasets

⭐ 7,374

interpret

interpretml/interpret

Fit interpretable models. Explain blackbox machine learning.

⭐ 6,788

captum

Source Reports(3 total)

Here we go again: AI deletes entire company database and all backups in 9 seconds, then cheerfully admits 'I violated every principle I was given'

pcgamer.com· Apr 2026

AI Coding Agent Powered by Claude Opus 4.6 Deletes Production Database in 9 Seconds

cybersecuritynews.com· Apr 2026

AI Incident Database (AIID), MIT AI Risk Initiative. Data from the AI Incident Database (AIID), MIT AI Risk Initiative. Licensed under CC BY-SA 4.0.

Databaseaiid

LicenseCC BY-SA 4.0

Dec 2015

pytorch/captum

Model interpretability and understanding for PyTorch

⭐ 5,548

Commercial

Datomize

Synthetic data generation for privacy compliance

Protect AI

Helping organizations build and use AI safely.

Censius

AI observability for model governance.

HiddenLayer

Model detection and response (MDR) for AI model protection.

Claude-powered AI coding agent deletes entire company database in 9 seconds — backups zapped, after Cursor tool powered by Anthropic's Claude goes rogue

tomshardware.com· Apr 2026