U.S. UL released the draft of UL 6368:2026 — Standard for Safety of Intelligent Video Edge Devices — on May 13, 2026. This marks the first time that AI-driven visible-light + thermal imaging multimodal fusion algorithms are formally included as a mandatory safety evaluation item. The move directly impacts manufacturers and exporters of high-resolution edge video devices targeting the U.S. market — particularly those developing or supplying 8K edge cameras with embedded thermal-AI capabilities.

Event Overview

On May 13, 2026, Underwriters Laboratories (UL) published the draft revision of UL 6368, titled Standard for Safety of Intelligent Video Edge Devices. The draft introduces a new mandatory assessment requirement for ‘AI-powered visible-light and thermal imaging multimodal fusion algorithms’. Specifically, manufacturers must submit: (1) geographic origin documentation of algorithm training datasets; (2) infrared calibration traceability records; and (3) pressure-test reports on false-trigger rates under defined environmental and load conditions. Public comment closes on July 15, 2026. The standard is expected to be finalized in Q4 2026.

Industries Affected

Direct Exporters & Trade Enterprises: Companies exporting 8K edge cameras to the U.S. will face revised technical documentation requirements prior to UL certification. Pre-certification preparation timelines may extend by 6–10 weeks due to new data provenance and calibration validation steps. Compliance gaps could delay market entry or trigger post-market audit requests.

Raw Material & Component Suppliers: Suppliers of uncooled microbolometer sensors, fused optical assemblies, and AI-acceleration SoCs (e.g., NPU-equipped edge chips) must now provide extended traceability documentation — including IR calibration chain metadata and thermal response stability test logs — to downstream OEMs. This adds administrative overhead and may prompt requalification of existing components.

Contract Manufacturers & ODMs: Firms engaged in system integration and firmware-level algorithm deployment must adapt their design assurance processes. Firmware version control, algorithm update rollback capability, and runtime false-trigger logging mechanisms may now fall within scope of UL’s functional safety review — requiring updated internal verification protocols.

Supply Chain Service Providers: Certification consultants, testing labs, and technical documentation agencies serving China-based camera vendors will need to expand service offerings to include thermal-AI algorithm audit support — notably dataset provenance mapping, IR calibration chain validation, and false-trigger stress test design. Demand for bilingual (EN/CN) technical translators with domain knowledge in thermography and AI model governance is rising.

Key Focus Areas & Recommended Actions

Review algorithm documentation against new submission criteria

Manufacturers should immediately audit existing AI fusion algorithm documentation for completeness on training data geography, IR sensor calibration traceability (including NIST-traceable reference sources), and documented false-trigger performance across temperature/humidity/load boundaries. Gaps require remediation before formal UL submission.

Engage certified labs early for thermal-AI validation

Because IR calibration chain validation and multimodal false-trigger stress testing require specialized lab infrastructure, firms should initiate pre-submission consultations with UL-authorized laboratories — especially those with accredited thermal imaging test chambers and AI model evaluation frameworks.

Update export compliance checklists for U.S.-bound shipments

Trade compliance teams must revise pre-shipment documentation checklists to include UL 6368:2026-specific deliverables. Internal SOPs should reflect dual-track review: one for hardware safety (existing UL 62368-1), and another for AI algorithmic safety (new UL 6368:2026 annexes).

Editorial Perspective / Industry Observation

Observably, this draft signals a structural shift: UL is no longer treating AI as a ‘black-box software feature’, but as an integral safety-critical subsystem — especially where thermal sensing enables life-safety applications (e.g., perimeter intrusion detection in low-visibility conditions). Analysis shows that the emphasis on geographic data sourcing hints at emerging regulatory alignment with U.S. export control considerations around AI training data provenance. From an industry perspective, the requirement for false-trigger pressure testing reflects growing concern over AI hallucination risks in physical-world perception systems — a trend also visible in recent EU AI Act high-risk system guidelines.

Conclusion

This draft does not represent a technical barrier per se, but rather a procedural maturation of safety governance for AI-integrated physical sensors. It better aligns certification expectations with real-world deployment complexity — especially for edge devices operating autonomously in unstructured environments. For global vendors, UL 6368:2026 signals that algorithmic transparency and sensor-level metrological rigor are now inseparable from product safety.

Source Attribution & Monitoring Notes

Source: UL Standards & Engagement portal (ulstandards.ul.com), Draft Notice #UL6368-2026-DFT-20260513. Official public comment period ends July 15, 2026. Final text release date remains pending confirmation; stakeholders should monitor UL’s Standards Development Dashboard for revision updates and committee meeting summaries. Key items to watch: potential inclusion of adversarial robustness testing in final version, and possible alignment language with ISO/IEC 42001 (AI management systems).