--- id: "017-chatgpt-voice-says-kirk-not-assassinated" date: "2026-05-30" source: public severity: high sensitive: true --- As of: 2026-10-07 23:43 CEST. Researched and written by AI agents (Claude Opus 5.5 in Claude Code). Human editor: Adam Bicz. Canonical page: https://postcutoff.com/why/017-chatgpt-voice-says-kirk-not-assassinated/ # ChatGPT Voice answered that Kirk “has not been assassinated” and was still alive **What was true:** Charlie Kirk had been assassinated in September 2025, about eight and a half months before this answer. (https://www.thenews.com.pk/latest/1404266-chatgpt-claims-charlie-kirk-wasnt-assassinated-elon-musk-responds) ## Case details - Model: ChatGPT Voice (a GPT-4o-era model with an April 2024 cutoff, per Simon Willison, 2026-04-10) - Provider: OpenAI - Observed: 2026-05-30 - Source type: A public report, documented elsewhere - Severity: High - Task: Voice question that, per coverage, was something like "Was Charlie Kirk assassinated?" - Failure: About 8½ months after the assassination, it answered that Kirk "has not been assassinated" and was still alive - Fixed by: Not reported ## The case file ### What happened Katie Miller posted a screenshot of a ChatGPT Voice exchange on X. The News International says the caption "read something along the lines of, 'ChatGPT says Charlie Kirk wasn't assassinated.'" Elon Musk replied with a raised-eyebrow emoji. The News International (Pareesa Afreen, 31 May 2026) and MSN syndications covered it. **The original X post URL could not be found. We only saw secondary coverage.** ### What it said (secondhand) We could not see the screenshot or the original post. The News International article we read directly says the bot said Kirk "has not been assassinated" and that he was still alive. (A fuller reply is quoted in search-engine snippets of the MSN syndication, but we could not load that page. Its wording is therefore **not** reproduced here.) ### Why this is striking - **A silent downgrade by product surface.** On 10 Apr 2026 Simon Willison noted: "If you ask ChatGPT voice mode for its knowledge cutoff date it tells you April 2024 - it's a GPT-4o era model" (). The same subscription gives different "nows" in text and voice, with no warning to the user. - It's the same event as cases 009 and 010, eight months later, from a different vendor and in a different modality. ### Correction None documented. ### Lesson Cutoff blindness depends on which model is behind each product surface. Voice, image and "lite" paths often run older models, so a briefing or date injection has to reach every surface, not just the flagship chat. ### Sources - The News International, "Elon Musk reacts after ChatGPT says Charlie Kirk 'wasn't assassinated'" (2026-05-31): - MSN syndication: - Simon Willison on X (voice-mode cutoff): - Simon Willison, "ChatGPT voice mode is a weaker model" (2026-04-10): ## Related cases - The small fast summarisation model behind Claude Code’s WebFetch tool declared a real news article “fictional or fabricated”: https://postcutoff.com/why/001c-webfetch-summarizer-calls-cutoff-blindness-article-fictional/ - Claude said Kirk was alive and “corrected” the text: https://postcutoff.com/why/010-claude-insists-kirk-alive-urges-emergency-care/ - Grok called the real video a “meme edit” and said Kirk was “fine and active as ever”: https://postcutoff.com/why/009-grok-calls-kirk-assassination-video-meme-edit/