Harvard Prof's AI Op-Ed Violates FT Rules; 3 Lawyers Cite 'Phantom' Cases

For AI practitioners, this story is not about politics — it is about the gap between model fluency and factual reliability. A Harvard professor's op-ed was publicly flagged after the Financial Times disclosed that 'AI was used to condense a longer draft,' while three attorneys apologized for Claude-generated 'phantom' quotations in a federal lawsuit. The common failure mode is hallucination meeting insufficient verification, and it is now producing public enforcement actions in both journalism and law.

A Harvard professor became the latest high-profile example of undisclosed AI use in professional writing after the Financial Times appended an editorial note to his op-ed attacking President Donald Trump's tariff policies, revealing that "AI was used to condense a longer draft of this column prior to submission to the FT and our own editorial involvement." The disclosure, reported on August 17, 2026, added that the FT's "editorial code of conduct specifically prohibits the use of AI in the writing process," and it followed social-media criticism of the column's "stilted prose." The unnamed professor was pilloried online in what two syndicated reports frame as political humiliation — but the more durable reading is a case study in the collision between generative AI and institutional editorial standards.

“

A Harvard professor's op-ed was publicly flagged after the Financial Times disclosed that 'AI was used to condense a longer draft,' while three attorneys apologized for Claude-generated 'phantom' quotations in a federal lawsuit.

The journalism episode does not stand alone. The same reporting connects it to a parallel failure in federal litigation, where three lawyers representing a former Department of Homeland Security official apologized to a California federal judge after one attorney used Anthropic's Claude Console to draft a motion containing "phantom" quotations — citations that looked authoritative but did not correspond to real cases. Attorney Greaves, who filed the motion on May 6, declared that he sent the AI-drafted document to an associate for fact-checking and that he reviewed the cited cases "generally" but did not verify each one. The underlying lawsuit challenges the Trump administration's mass layoffs of federal employees and seeks to depose Joseph Guy, the former DHS deputy chief of staff under Secretary Kristi Noem, over cuts at the Federal Emergency Management Agency.

The reporting also notes this is not the first time anti-Trump actors have enlisted chatbots: a lawyer was previously forced to apologize to a federal judge for using AI to draft a lawsuit against the administration's federal workforce reductions. The syndicated pieces frame these as a partisan pattern, but the more useful technologist's reading is that AI misuse is spreading across domains because the tools are good enough to pass cursory review while still capable of embedding fabricated supporting detail.

Both incidents share a single failure mode. Large language models produce fluent, confident prose that can contain subtle fabrications, and time-pressed professionals are delegating verification to subordinates — or skipping it — while treating AI output as a finished product. For the AI industry, the reputational damage is concrete: Anthropic's Claude is named directly in the legal case, attaching a frontier enterprise model to a public hallucination incident, while the FT's note shows a prestige publisher willing to publicly name and shame AI-assisted submissions rather than quietly correct them. For Harvard, the professor's affiliation turns an individual lapse into an institutional reputational event, underscoring how elite brands are now exposed to a single author's verification habits.