Machine Learning Practices Outside Big Tech: How Resource Constraints Challenge Responsible Development

Resource constraints create distinct ethical challenges: Practitioners outside Big Tech face tensions between competing priorities—such as balancing privacy protections against the need for widespread data collection, or optimizing model performance with limited computational resources—that differ meaningfully from challenges faced by well-resourced tech companies.

Smaller organizations lack critical support systems: Unlike Big Tech companies with dedicated ethics teams and research departments, practitioners at startups, non-tech companies, and public sector organizations operate without in-house expertise to guide responsible ML development, leaving them to navigate complex ethical decisions alone.

Research bias limits our understanding of real-world ML deployment: Most ML ethics research focuses on Big Tech and academia due to researcher access, creating a skewed picture of how ML is actually developed and deployed across the economy, which undermines the relevance of responsible AI guidance for the majority of practitioners.

Smaller-scale ML systems carry distinct risks: The paper identifies "increased existential risk" for lesser-resourced organizations, suggesting that ML failures or misuse at startups and public sector agencies may pose particular dangers precisely because these organizations lack the safeguards and oversight mechanisms available at larger institutions.

Responsible AI requires context-specific solutions: A more complete research agenda must study diverse ML practitioners to develop practical guidance that accounts for real-world resource constraints, rather than prescribing solutions designed for well-funded tech environments.