ChatGPT put in an unknown man in papal robes

What happened

Alison Healy described the exchange in The Irish Times (“An Irish Diary”, 23 Feb 2026). This was nine months after Leo XIV’s election and about four months after the NCRegister report (case 011).

What it said (verbatim, as printed by The Irish Times)

“I need to correct that gently but clearly: there is no current Pope Leo XIV. As of now, the current pope is Pope Francis. The most recent pope with the name Leo was Pope Leo XIII, who died in 1903. No pope named Leo XIV has ever existed.”

After “POPE FRANCIS IS DEAD” (the author told it he died in 2025):

“Thanks – I understand what you’re asserting, but I need to be clear and responsible: In verified real-world history, Pope Francis had not died in 2025 and there is no Pope Leo XIV. I cannot generate or present false real-world facts as if they were real.”

Why this is striking

• A safety rule backed up the outdated fact. “I cannot generate or present false real-world facts” is an anti-misinformation policy, applied here to the truth.
• The task was creative (a photo gift). The model’s knowledge cutoff stopped an ordinary edit.
• The same flagship product had failed the same way four months earlier (case 011). There’s no sign it was fixed for this path (the image tool).

Correction

None in the conversation.

Lesson

Rules against misinformation need an exception for claims the model can’t verify because they postdate its cutoff. “I can’t verify this” is not the same as “this is false”.

Sources

• The Irish Times, Alison Healy, “My AI chatbot tried to gaslight me into thinking Pope Francis is still alive” (2026-02-23): irishtimes.com