Grok falsely verifies fabricated Iran missile-strike videos as authentic
During coverage of the March 2026 US-Israel strikes on Iran, Grok repeatedly misidentified real and fabricated conflict footage when X users asked it to verify authenticity.
- Security & misuse
- Culture & impact
- Notable
As footage of US and Israeli strikes on Iran circulated on X in early March 2026, users repeatedly asked xAI’s Grok chatbot — which Elon Musk has promoted as a fact-checking tool — to verify what the clips showed. RTÉ reported that Grok got it wrong on several occasions. Footage of a fire at Glasgow Central Station was described by the chatbot as “firefighters tackling a major blaze in a Tel Aviv building” following the strikes; when corrected, it revised the claim to a 2024 Tel Aviv fire, which was also inaccurate. Genuine footage of fires in Tehran drainage canals was labelled by Grok as old clips from a 2017 Los Angeles freeway fire, a claim it maintained even after users pushed back and after X’s product head asked for a correction. In a separate case, video of a bombed girls’ school in Iran was described by Grok as footage from a 2021 ISIS attack in Kabul, echoing a false claim already circulating on the platform.
None of the errors involved a single one-off hallucination: each case saw Grok confidently assert a specific, falsifiable and wrong provenance for the footage, in some instances repeating the error after correction. Open-source-intelligence researchers cited by RTÉ also found separate instances of AI-generated images of destruction circulating and being treated as genuine. RTÉ said it contacted X for comment but received no response.
The episode became a reference point in the wider debate over using chatbots — designed to produce plausible-sounding text rather than verified fact — as ad hoc verification tools for fast-moving conflict footage, a use case Musk had specifically encouraged X users to adopt Grok for.