Vertu’s $6,880 AI phone shows why agent autonomy needs better guardrails
A test of Vertu’s luxury AI phone highlights the trade-off between giving mobile agents more authority and keeping their actions reliable.
Vertu’s Alphafold smartphone is built around a preinstalled assistant called Hermes Agent and starts at $6,880. The device is designed to analyse documents, perform actions across applications, plan travel and escalate requests to a human concierge.
What happened
Independent testing found that Hermes was more willing than mainstream assistants to perform multi-step actions autonomously. However, it also created calendar entries for incorrect dates, produced inaccurate reminder times and behaved inconsistently when accessing previously uploaded files. The handset itself was developed through a supply-chain partnership involving established smartphone hardware, while Vertu provides the materials, software and service layer.
Why it matters
AI-phone competition is moving from conversational assistants toward agents that can act inside applications. That makes reliability more important. A hallucinated answer is inconvenient, but an incorrect calendar entry, message or transaction can have direct consequences.
The bigger picture
The device illustrates the central tension in agentic products: greater autonomy can make an assistant more useful, but it also expands the cost of mistakes. Consumer adoption will depend on permission controls, confirmation steps, audit trails and predictable recovery when an agent acts incorrectly. Vertu’s price makes it a niche product, but the underlying design problem applies to the entire mobile-agent market.