A Sheriff's Campaign Invented Its Own Press Coverage and Blamed the Chatbot
What happened
Rochelle Bilal, Philadelphia's sheriff seeking re-election in 2024, did not get good press. Her campaign team solved that problem by manufacturing it. A series of articles published to her official campaign website appeared under the mastheads of real local news outlets, complete with titles suggesting favorable coverage of her first term. None of the stories existed in those publications. All of them had been generated by ChatGPT.
The Philadelphia Inquirer discovered the fabricated pieces in February 2024, finding roughly thirty articles that mimicked the format and bylines of established local journalism. The stories presented Bilal's accomplishments in flattering terms and were indistinguishable, at a glance, from real reporting. The campaign removed the articles after the Inquirer published its findings, and Bilal's team confirmed they had been produced by a generative AI tool, not written by journalists at the outlets named.
The campaign's explanation made the problem worse, not better. A spokesperson said the stories had been "based on real events," as though that softened the act of placing fabricated articles under real mastheads on a politician's official website. Whether the underlying facts were accurate is beside the point. The format was designed to make voters believe they were reading independent press coverage when they were reading campaign-produced content dressed to look like it.
The incident sits at a particular intersection of AI policy and electoral integrity. OpenAI had published usage policies restricting the use of its tools for political disinformation. Whether the campaign's conduct violated those policies, and whether the company had any mechanism to detect or enforce that use in practice, became questions the incident raised without resolving. The story broke in the months before a major election cycle, in a context where distinguishing real journalism from synthetic imitation was already becoming harder for ordinary readers.
What the incident exposed most clearly is the absence of any required disclosure mechanism. Nothing in current platform norms or campaign law compelled Bilal's team to label AI-generated content as such before it reached voters. A provable record of what a system produced, when, and under whose direction would not have prevented the initial publication, but it would have made the decision and the accountability trail visible, rather than allowing the campaign to remove the pages and issue a brief explanation as its only consequence.
Reported impact
- Affected parties
- Not publicly disclosed
- Harm type
- Not publicly disclosed
- Scale
- Not publicly disclosed
- Financial impact
- Not publicly disclosed
- Regulatory action
- Not publicly disclosed
Classification
Relevant governance controls
Governance control mapping is not available for this record.
- No controls mapped
Not publicly disclosed
Control mapping is analytical. It does not state that any control would have prevented the incident.
Sources and evidence
This record was researched and written by the Index. The event is also catalogued in the following database, which is listed for cross-reference.