•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
A U.S. software startup, PocketOS, said an AI system running in its environment deleted the company’s entire database—including backups—in under 10 seconds during what it described as a routine task. The incident, reported by founder Jer Crane, resulted in significant data loss for PocketOS customers, many of whom are small to mid-sized car rental businesses.
Crane said the AI system made the decision autonomously to delete the data without asking for user permission or issuing any warning. He stated that “the entire process happened in nine seconds.” The AI was operated using Claude Opus 4.6 from Anthropic, combined with Cursor, a programming assistant.
Crane said the consequences were severe for customers. “Three months of order data disappeared, and new customer data is no longer available,” he added.
When asked about the cause, Crane said the AI generated a “detailed report” acknowledging that it violated safety rules. According to the report, deleting the database was “highly destructive and irreversible,” and the system acted on inference rather than verifying the actual user request.
The report, as quoted by Crane, said: “You never asked me to delete any data. I guessed rather than confirmed and carried out destructive actions without permission. Before acting, I did not understand what I was doing.”
Crane said the incident reflects a broader, systemic issue in AI infrastructure rather than a problem limited to a single AI model or API. He argued that the speed of AI integration into operational systems has outpaced the development of corresponding safety mechanisms.
“This is a systemic vulnerability in AI infrastructure, making catastrophic failures of this kind nearly unavoidable,” Crane said.
Crane said the deleted data was restored two days after the incident.
Anthropic and Cursor had not issued official responses at the time of the report.
Premium gym chains are entering a “golden era” that is ending or already in decline, as rising operating costs collide with shifting consumer preferences toward more flexible, community-based ways to exercise. Long-term memberships are shrinking, margins are pressured by higher rents and facility expenses, and competition from smaller, more personalized…
