Dale Caldwell’s resignation as New Jersey’s lieutenant governor has moved from a political and ethics story into a test case for how public figures talk about AI. After an investigation found he had sexually harassed a staffer and repeatedly violated ethics rules, Caldwell has argued that AI tools support his view that he is being unfairly targeted.
What Caldwell says AI found
Caldwell was forced to resign on September 25th after the investigation’s findings. Since then, the now-former Lt. governor has been appearing in the media as he tries to defend himself and clear his name.
In an interview on NJ PBS, Caldwell described using artificial intelligence to review the report on the investigation. He told host Rob Nelson that he “AIed” the report “from multiple AI platforms.”
“I put it through AI,” he said, “and it said 59 times, it [sic] said, ‘What would your findings be?’ There was no instance in any of the AIs that said there would be a finding of sexual harassment.”
That statement puts AI at the center of his defense. Caldwell is not only disputing the findings; he is pointing to outputs from multiple AI systems as support for his position.
Why the AI claim matters
The claim is notable because AI chatbots are not courts, investigators, or ethics panels. They generate responses based on prompts, context, and patterns in data. The source article notes that AI chatbots are notoriously sycophantic and often tell users what they want to hear.
That matters in a case where the user has a strong personal stake in the answer. If Caldwell asked AI systems to assess the report, the result may depend heavily on how the report was entered, what question was asked, and what framing surrounded the request.
The central issue is not whether AI can summarize documents or help a person think through arguments. The issue is whether chatbot responses should be treated as meaningful proof against an investigation’s findings. Based on the source, Caldwell is presenting AI outputs as part of his public argument, but the article gives no indication that those outputs replace the investigation itself.
The gap between analysis and authority
AI tools can appear confident even when their answers are shaped by incomplete context or leading prompts. That makes them especially risky in high-stakes disputes involving accusations, employment conduct, ethics rules, and public office.
In Caldwell’s case, the investigation found two categories of misconduct:
- Sexual harassment of a staffer
- Repeated violations of ethics rules
Caldwell’s AI argument focuses on whether the systems would have found sexual harassment. But the source article does not say that AI reviewed the matter with official authority, conducted interviews, assessed credibility, or held any formal role in the investigation.
That distinction is important. An AI-generated response may sound like an assessment, but it does not carry the same weight as an investigative process. It also does not explain how the systems reached their answers, whether they had the complete record, or whether the prompt shaped the conclusion.
A broader warning for public debates
Caldwell’s comments show a pattern that may become more common: people using AI outputs as a rhetorical shield. Instead of saying only that they disagree with a finding, they can say that “multiple AI platforms” agree with them.
That phrase can sound technical and neutral. But without the prompts, inputs, and full outputs, readers cannot evaluate what the AI actually did. The number “59 times” may sound precise, yet it does not by itself establish independence, accuracy, or fairness.
The source article also points to a practical problem: if a user wants a certain answer, a chatbot may produce responses that reinforce that preference. That does not mean every AI response is useless. It does mean AI output should be handled carefully, especially when someone is using it to contest serious findings.
For readers, the useful question is simple: what is being verified, and by whom? In this case, the reported facts are that Caldwell resigned after an investigation found sexual harassment and repeated ethics violations, and that he later said AI systems did not support a sexual harassment finding. Those are different kinds of claims. One comes from an investigation; the other comes from Caldwell’s description of chatbot responses.
What remains clear
The public record described in the source is narrow but consequential. Dale Caldwell was New Jersey’s lieutenant governor. He was forced to resign on September 25th after an investigation found he had sexually harassed a staffer and repeatedly violated ethics rules. He has since argued publicly that he was unfairly targeted.
His use of AI does not erase the investigation. It does, however, show how quickly AI can become part of reputation management, political messaging, and disputes over accountability.
As AI becomes easier to use, public figures may increasingly cite chatbot outputs in their own defense. Caldwell’s case shows why that should be treated with caution. A chatbot answer can be a talking point, but it is not the same thing as evidence, process, or independent judgment.