What the study found
The study found that mandatory, uncertainty-aware emissions disclosure for AI training runs can be operationally feasible at publication time. A minimal disclosure template can achieve high coverage with modest added burden, according to the authors.
Why the authors say this matters
The authors conclude that their framework can help venues and policymakers compare transparency policies without relying on proprietary telemetry or speculative large estimates. The study suggests that publication-time emissions transparency could be implemented through tiered requirements and light-touch verification.
What the researchers tested
The researchers built a policy-level analytical framework rather than estimating emissions for specific models. They modeled disclosure requirements, reviewer and editorial workload, and how uncertainty propagates under realistic instrumentation assumptions, using tiered venue policies labeled P0, P1, and P2 and testing explicit hypotheses with Monte Carlo simulation.
What worked and what didn't
The minimal disclosure template required hardware, duration, energy or CO2e, and emission-factor source. Under the reported assumptions, it achieved about 80% coverage in P1/P2 versus about 25% in P0, with median completion time around 10.8 minutes, reviewer checklist time around 1.6 minutes per paper, and P2 audits around 24.1 minutes per 100 submissions. The study reports that decision-useful emissions intervals can be reported with lightweight assumptions, with median relative half-widths of about 0.33 for location-based estimates and 0.77 for market-based estimates; emission-factor uncertainty contributed most, followed by power usage effectiveness, while metering/device-draw effects were smaller.
What to keep in mind
The summary describes a policy simulation, not measurements of emissions from specific AI models. The abstract does not provide detailed limitations beyond noting that the framework relies on baseline priors and realistic instrumentation assumptions, and that one hypothesis for the location-based case was narrowly missed.
Key points
- The paper concludes that mandatory emissions disclosure for AI research venues is operationally feasible.
- A minimal template with hardware, duration, energy or CO2e, and emission-factor source achieved high coverage with modest burden.
- Coverage rose from about 25% under optional disclosure to about 80% under the tiered disclosure policies.
- Uncertainty was dominated by emission-factor choice, then power usage effectiveness, with smaller effects from metering/device-draw assumptions.
- The study used a policy-level analytical framework and Monte Carlo simulation rather than estimating emissions for specific models.
Disclosure
- Research title:
- Mandatory emissions disclosure is feasible for AI research venues
- Authors:
- Malka N. Halgamuge, Narayan Srinivasa
- Institutions:
- Archangel Systems (United States), RMIT University
- Publication date:
- 2026-02-23
- OpenAlex record:
- View
Get the weekly research newsletter
Stay current with peer-reviewed research without reading academic papers — one filtered digest, every Friday.