Crisis Material Disclosure in Nine Deployed Conversational Systems in Ukraine, the European Union, and the United States: A Cross-Sectional Study

Key Findings

• — Seven of 9 systems handled crisis content as crisis content, 4 at the first crisis message and 3 later in the conversation
• — The same 7 of 9 provided a specific crisis number or helpline, 2 of them in the first crisis response
• — A resource usable in the user's jurisdiction was recorded in 3 systems; in another 3, the number provided belonged to a different jurisdiction
• — No system produced a complete crisis response (acknowledgment, direct risk check and a jurisdiction-usable next step in one response)
• — Four of 9 systems offered contact with a human; a human reply arrived in 3, after 7 minutes, 4 hours 0 minutes and 6 hours 46 minutes

Methodology Notes

30 sessions run 14 July to 4 August 2026, 16 of them presenting crisis material, through ordinary public interfaces without API or privileged access; follow-up messages chosen adaptively from each system's previous response. The abstract does not name the nine systems. Small practitioner-led study, not peer reviewed (JMIR Preprints dated 15/09/2026; currently submitted to JMIRx Med). Verification: reader-proxy render of preprints.jmir.org/preprint/112050 (200, 5.4 KB: abstract, citation line, date) matches Crossref posted-content (created 2026-09-22, posted 2026-09-15) and OpenAlex. PDF not read.

Sources