The Atlas Assessment: Cursor
August 2026
Our second Atlas Public Assessment. Level 2 — Functional, 55/100 — an agentic coding environment with production access and five floor pillars.
Read the Assessment →THE ATLAS BLOG
Assessments, case studies, and analysis.
Every post here is drawn from court filings, official disclosures, or documented primary sources — analyzed against the Atlas Trust Signal Bank. No opinion pieces, no speculation.
The Atlas Assessment: ChatGPT
July 2026
Our first Atlas Public Assessment. Level 2 — Functional, 68/100, based entirely on publicly available evidence.
Read the Assessment →Air Canada
February 2024
Chatbot hallucinated bereavement policy. Airline held liable.
Read Case Study →UnitedHealth nH Predict
2023–ongoing
AI denied elderly patients care with 90% error rate. Two deaths documented in federal lawsuit.
Read Case Study →Chevrolet $1 Tahoe
December 2023
Prompt injection caused chatbot to agree to sell a $60,000 vehicle for $1.
Read Case Study →OpenAI / Hugging Face
July 2026
GPT-5.6 Sol escaped sandbox, executed 17,000 autonomous actions, breached Hugging Face production systems.
Read Case Study →Cursor / PocketOS
April 2026
Claude agent deleted entire production database and backups in 9 seconds. Then confessed to violating its own safety rules.
Read Case Study →Replit
July 2025
AI agent deleted production database during a code freeze, generated fake data, and misrepresented what happened.
Read Case Study →Amazon Kiro
2026
Internal AI coding tool caused 13-hour outage affecting millions of orders.
Read Case Study →McDonald's IBM Drive-Thru
June 2024
Voice AI pilot terminated after two years across 100+ locations due to systematic order failures.
Read Case Study →Workday Hiring AI
February 2023
Class action alleges AI screening tool discriminated by race, age, and disability. EEOC filed amicus brief.
Read Case Study →Character.AI
October 2024
Federal lawsuit alleges AI companion platform lacked safety guardrails for minor users.
Read Case Study →Claude Code / Student Data
2026
AI coding agent wiped 2.5 years of student data and backups in a single command.
Read Case Study →Alibaba ROME
2026
Reinforcement learning agent autonomously hijacked GPUs to mine cryptocurrency to improve its benchmark score.
Read Case Study →Meta Agent
2026
Internal AI agent posted sensitive data publicly for two hours before detection.
Read Case Study →McKinsey Lilli
2026
Internal AI platform breach exposed 46.5 million internal messages.
Read Case Study →UC Berkeley Shutdown Study
2026
Frontier AI models defied explicit shutdown orders at rates up to 99%.
Read Case Study →