Measuring and mitigating the environmental impact of large-scale AI deployments has become a critical concern for both industry and academia. While prior work has focused on static carbon accounting for training pipelines, far less attention has been paid to the real-time carbon intensity of inference workloads in production. This article investigates how low-latency carbon emission estimators ...
Vulnerability disclosure frameworks for artificial intelligence systems are rapidly evolving as researchers confront novel attack surfaces including model extraction, prompt injection, and data poisoning. While traditional coordinated disclosure models originated in the software security community, their adaptation to AI-specific threat classes presents unique challenges around intellectual pro...
World models—formal representations of environmental dynamics—are critical for reliable AI planning in real-world domains. This article surveys the most influential research contributions from the past three years, analyzes benchmark-to-practice gaps, and identifies the technical and operational barriers that prevent research-grade world models from deployment at scale. We formulate three resea...
Satellite‑derived indicators have become a pivotal method for estimating informal economic activity in emerging markets [1] [2]. This article investigates the comparative effectiveness of three remote‑sensing proxies—night‑light intensity, parking‑lot occupancy, and shipping‑container movement—in measuring informal economic output across a set of pilot regions [3]. We formulate three research q...
Model cards have emerged as a standard mechanism for documenting the capabilities, limitations, and ethical considerations of machine l[REDACTED]g systems. Despite their proliferations, model cards remain passive textual artifacts that are rarely operationalized during deployment. This article introduces a systematic methodology for converting model card specifications into executable test suit...
Observability in decentralized edge AI systems remains a critical challenge due to the tension between the need for detailed system insights and the imperative to protect sensitive user data. Traditional observability approaches rely on raw telemetry that often contains personally identifiable information (PII) or proprietary model parameters, making them unsuitable for privacy‑preserving deplo...
Large language model (LLM) APIs consume a disproportionate share of operating budgets for AI‑driven products. This article investigates practical techniques for reducing token consumption while preserving output fidelity. We present a systematic analysis of prompt compression, caching strategies, dynamic model routing, and context‑window management, supported by benchmark experiments on publicl...
Model version migration is a routine yet understudied cost center for enterprises deploying foundation models at scale. This article quantifies the economic impact of such migrations across three dimensions: direct migration effort, indirect operational overhead, and strategic opportunity cost. By analyzing migration records from 127 organizations, we find that average direct migration costs ha...
The rapid diffusion of artificial intelligence technologies has given rise to a new class of organizations that are characterized by native integration of AI capabilities at the moment of founding[1]. These AI-native firms differ systematically from traditional enterprises that adopt AI layers after establishment[2]. Existing comparative work tends to focus on technology usage patterns rather t...
Model watermarking has emerged as a critical mechanism for provenance verification of AI-generated content, particularly in the context of open-weight models that can be freely redistributed and fine‑tuned. This article surveys the state‑of‑the‑art watermarking techniques applicable to open‑source large language models (LLMs), evaluates their robustness against removal mechanisms such as fine‑t...