Knowledge at Wharton Ripple Effect: What We Know About AI Fighting Corruption | Philip Nichols
Nov 4, 2025
Philip Nichols, a Professor of legal studies and business ethics at the Wharton School, dives into the complexities of AI in the fight against corruption. He explains that current AI models are not yet equipped to detect corruption effectively due to data limitations. Nichols discusses regulatory differences between the EU and the U.S. that impact AI use. While AI shows promise in identifying red flags in large datasets, he warns about its potential misuse by criminals and stresses the need for human oversight in any anti-corruption strategy.
AI Snips
Chapters
Transcript
Episode notes
Corruption Is Heterogeneous And Data-Poor
- Corruption varies widely across countries, industries, and firms, so data about it is not fungible.
- AI needs large, consistent data and a solid model, which corruption detection often lacks, causing hallucinations.
AI's Data And Model Limits Matter
- Current AI requires lots of data and accurate models, which compliance contexts often lack.
- Without them, AI produces hallucinations or nonsense that can harm people and processes.
Regulatory Regimes Shape AI Use
- Regional AI rules differ: EU emphasizes individual dignity while U.S. rules are looser.
- Transnational firms must follow stricter EU standards or risk infringing worker rights when monitoring employees.