adversarial archives

The adversarial archives tag covers discussions about how curated, partisan document collections are used to probe generative AI systems, particularly in the context of the long-running Donovan–Shell dispute. Threads describe experiments where decades-old archives about Royal Dutch Shell were fed into multiple public AI assistants, producing divergent and sometimes contradictory outputs. These tests expose weaknesses in model grounding, provenance, and corporate reputation management, as AI hallucinations and conflicting narratives emerge from the same source material. The tag focuses on the intersection of contested corporate history, AI behavior, and the strategic use of archives to challenge institutional narratives.