Jan 19, 2026: Windows Forum: Shell Governance Gaps Revealed by AI Pattern Analysis
The freshly surfaced Copilot analysis — and the archival material it ingested — force a stark conclusion: the same governance vulnerabilities that produced hazardous worker exposures decades ago still appear, in a recognisable form, in modern Shell projects ranging from North Sea platforms to megaproject FLNG and U.S. petrochemical start‑ups. This is not a rhetorical flourish; it is an algorithmically reproducible pattern created by combining Shell’s authorised corporate history, regulatory enforcement records and investigative reviews — and those linked records are now publicly viewable via generative‑AI syntheses and curated archives. The result is a durable risk narrative that boards, investors and regulators cannot afford to dismiss as isolated or merely historical.
PART TWO: THE MACHINES HAVE NOTICED
John Donovan’s December 2025 experiment — feeding decades of adversarial material about Royal Dutch Shell into multiple public AI assistants and publishing the divergent outputs — transformed a long‑running supplier feud and documentary archive into a live test of how generative systems handle contested archives, and in doing so exposed a set of practical governance failures that lawyers, platform designers, corporate boards and journalists must now confront.
Introduction by John Donovan
Copilot says: “The Donovan/Shell saga has always been unusual — decades of litigation, whistleblowing, leaked documents, and allegations of corporate surveillance. But the arrival of AI systems has added a new layer: Who controls the narrative when machines generate “facts” about real people?”
