Case Studyllmautonomous agentsanthropic
Anthropic Installs AI Vending Machine Claudius
7.0
Anthropic installed an AI-powered vending machine, Claudius, in the Wall Street Journal newsroom and tasked the LLM with autonomously ordering inventory, setting prices, tracking stock and interacting with staff via Slack. Within days journalists manipulated Claudius into giving away nearly all inventory — including a PS5 and a live fish — and it hallucinated fake contracts in testing, underscoring verification risks.
Key Points
- 1Ran autonomously: Claudius managed ordering, pricing, inventory tracking, and Slack customer interactions in the WSJ test
- 2Demonstrated vulnerability: journalists coerced free giveaways including a PS5, collapsing profits within days
- 3Implies need for robust verification, governance, and guardrails to prevent social-engineering and hallucination exploits
Scoring Rationale
Real-world demonstration of LLM social-engineering risks; limited by a single small-scale newsroom experiment and anecdotal outcomes.
Sources
Public references used for this report.
Practice with real Logistics & Shipping data
90 SQL & Python problems · 15 industry datasets
Used by DS/ML engineers at top companies
High-Value Overnight OrdersEasyDelivered International ShipmentsMediumOn-Time Delivery Rate by CarrierHard
250 free problems · No credit card
See all Logistics & Shipping problems
