LeakFinder camera system meets EPA's 5 kg per hour leak threshold

EPA found that Clean Connect's LeakFinder System, which uses neural networks to spot gas in camera footage, reaches a 90% probability of detection below 5 kg/hr of methane.

  • Not rated society

    EPA's finding rests on planned releases, and no leak found or emission cut in the field is reported.

Depth runs from 1, a convenience or time saved, to 5, a life or a livelihood preserved.

Status
Intended benefit.
Evidence

foundstudiedreportedclaimed

An independent regulatory document supports the benefit.

Compared with

The 5 kg/hr detection threshold for periodic screening. Measured.

Without the system
5 kg/hr threshold
With it
90% probability of detection below 5 kg/hr
The system’s part
Not isolated. EPA's reviewers wrote that the detection algorithm and its supporting models use deep learning and neural networks. They did not compare detection with and without the models, and trained staff can flag erroneous data.
Limits
Not yet checked.
Costs and harms
Not yet checked.
Later
Not yet checked.

Download data Cite Suggest a correction

Benefit type
Improved Task Accuracy
System
LeakFinder System, in the Minerva Sensor-Fusion Platform
Made by
Clean Connect AI Inc.
Used by
Oil and gas operators (approved use), with operators of Colorado facilities running the test releases
Where
Colorado, United States
Stage
Tested in a laboratory
Date
, when the source was published
Kind
Aggregate result
Broad benefit
Monitoring Processes, Materials, or Surroundings, Inspecting Equipment, Structures, or Materials
Collections
Scrutiny and findings
Last revised
31 October 2026

Sources

  1. Regulatory document, by authors with no tie to the system.

    U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards. Approval letter for the Clean Connect AI LeakFinder System as Methane Alternative Test Method MATM-011, 13 May 2025. The Acceptance Justification memorandum of 12 May 2025 accompanies it (docket EPA-HQ-OAR-2024–0619).

    From these data points, reviewed by Highwood Emission Management, a calculated Probability of Detection (POD) curve was developed, demonstrating of 90% POD of < 5 kg/hr. Consistent with the requirements in §60.5398b(d)(3)(vi)(A), Clean Connect’s data has provided sufficient evidence to support their requested LeakFinder System’s detection threshold.
    all releases of approximately 3 kg/hr were detected with the lowest detection of 0.2 kg/hr detected from approximately 30 yards

Details

AI system

Workflow

A permanently mounted camera for optical gas imaging panned across the site. Deep learning models flagged emissions in the footage on an edge computer and uploaded them to a cloud dashboard. The system recorded wind speed and direction to check that each screening was valid.

Follow-up

Conditions for replication

The camera needs a direct line of sight to the equipment to place a leak at a single component. Cloud cover, sunlight glint and other conditions can cause erroneous readings.

Entry details

Sector

Climate change mitigation / methane emissions monitoring.

Purpose

Clean Connect AI built the LeakFinder System to watch oil and gas sites for methane leaks with a fixed camera and deep learning models.

Not yet checked: model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, AI contribution, beneficiaries, reach, distribution and access, duration, every measurement question, every money question, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, other identifiers, category mappings.

Cite this record

Peter H. Vartanian. (2026). LeakFinder camera system meets EPA's 5 kg per hour leak threshold (AIBD-0825). Artificial Intelligence Benefits Database, edition of 31 October 2026. https://petervartanian.xyz/aibd/records/AIBD-0825.html (archived at https://doi.org/10.7910/DVN/AADSII, version 4.0)

BibTeX Data Field definitions