From Cradle to Cloud: A Life Cycle Review of AI's Environmental Footprint
2026-05-08T07:23:57Z•01061ff6618eec46a006dd129f983449b97bf0537e3d1e289824a3f81859b61b
AI environmental footprintAI interpretabilityCO2e reportingDigital Services ActEU AI ActGDPRLLM social effects','AI tutoring evaluation','student behavior',LLMorphismPrivacy Access TokensPrivacy PassStackelberg gameaudit gamingauditor policiescontinuous compliancecryptographic age verificationdriver monitoring systemshuman cognition biasin-cabin sensinglife cycle assessmentmodel evaluationmodel training and inferenceprivacyprivacy-preserving age verificationregulatory evasionsimulator
What happened
This collection of recent CS papers highlights recurring security, privacy, regulatory and socio-technical risks across AI life cycles and deployments. Key findings: (1) AI environmental-impact assessments lack standardized life-cycle metrics and over-rely on coarse CO2e proxies, hindering comparability and policy action; (2) continuous post‑deployment compliance (e.g., under the EU AI Act / DSA) enables novel strategic gaming by regulated systems (delay, drift, cherry‑pick, attrition, OffAuditDrift) and requires history‑aware audit policies and redesigned monitoring to close coverage/granular
Why it matters
A reviewed impact interpretation has not been published for this record.
Evidence and limitations
- Source ID
- arxiv_cs_cy
- Record identifier
- 01061ff6618eec46a006dd129f983449b97bf0537e3d1e289824a3f81859b61b
- Enrichment time
- 2026-05-08T07:23:57Z
- AI-assisted enrichment
- Yes
This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.