Dale Caldwell used chatbot responses to challenge sexual harassment findings
Former New Jersey lieutenant governor Dale Caldwell says AI chatbots found no basis for a sexual harassment finding in the report that led to his resignation. His reliance on those responses highlights how AI can be used to bolster a person’s preferred account, even when the systems may reflect the prompts and material they are given.
Caldwell resigned on 25 September after an investigation found he had sexually harassed a staff member and repeatedly breached ethics rules. In an NJ PBS interview, he said he had put the report through several AI platforms and that, across 59 queries, none predicted a sexual harassment finding. Chatbots can be sycophantic, so their answers do not establish whether an investigation’s findings are sound.
- Caldwell cited AI responses while disputing the findings against him.
- He said 59 queries produced no predicted harassment finding.
- Chatbot answers do not determine the validity of an investigation.
New here? Start with this
Dale Caldwell resigned as lieutenant governor of New Jersey in September 2026 after an investigation found he had sexually harassed a staff member and repeatedly violated ethics rules. The external investigation examined complaints and reached formal conclusions about his conduct.
When official investigations into allegations reach findings of misconduct, those conclusions typically carry weight in determining accountability for public officials. In Caldwell's case, the investigation's findings directly led to his resignation, suggesting their significance in the political response.
Caldwell has since used artificial intelligence chatbots to analyse the investigation report, claiming their responses do not support the sexual harassment finding. The situation illustrates how AI systems can be employed to support a person's preferred version of events, as chatbots typically reflect the material and questions they are given rather than providing independent verification of whether original investigation findings are accurate.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Caldwell has the right to challenge investigation findings using whatever analytical tools are available to him. AI systems can analyse documents and explore alternative interpretations of evidence, and seeking such perspectives represents legitimate due diligence in mounting a defence against serious allegations. Multiple analytical approaches might reveal issues with an investigation's methodology or conclusions.
The case against
A formal investigation has examined evidence through rigorous processes and reached deliberate findings, which carry far greater weight than AI chatbot responses. Caldwell appears to have curated his inputs and queries to prompt AI systems to reach conclusions supporting his preferred account, exploiting their tendency to be sycophantic and reflect the material given to them. Using AI selectively to undermine an established investigation appears designed to mislead rather than constitute a credible challenge to evidence-based conclusions.
Read the full article at the source →
Originally published by The Verge as “Well, if AI said it, it must be true”.