Amazon and Google are aggressively acquiring unconventional data sources to train AI models: Amazon covertly bulk-purchases rare out-of-print books through intermediaries, systematically scanning and destroying them to fill knowledge gaps; Google paid $10M for Spirit Airlines' operational data at bankruptcy auction including internal wikis, emails, and source code, despite labor union protests over privacy risks. Both cases highlight the intense competition for training data and emerging ethical/legal gray zones around data acquisition that Thai tech companies should monitor.