Grok misidentifies video of US military exercisepublished at 14:15 GMT

14:15 GMT

Thomas Copeland
BBC Verify Live journalist

The artificial intelligence chatbot Grok, which is owned by social media firm X, has wrongly characterised a video from a US Marine flypast in California as being of a US military exercise near Venezuela.

The footage shows a large number of military aircraft flying in formation down a beach and has spread widely online.

One post by the Tehran Times, a newspaper considered to be heavily supportive of the Iranian government, gathered more than a million views on X and falsely claimed it shows a “US military exercise just off the Venezuelan border”.

When asked by X users if this was true, Grok repeatedly confirmed the false claim, even incorrectly insisting that specific military aircraft are visible in the footage.

The video is actually from a 250th anniversary celebration and live fire demonstration by the United States Marine Corps on 18 October.

Graphic with a screenshot of the video showing at least six helicopters and 10 fighter jets flying over a beach. Nexto to it is a post on X from Grok labelled "false", saying: ""The video depicts a recent U.S. Air Force exercise off Venezuela's coast, involving B-52 or B-1 bombers in a simulated attack demonstration. This aligns with Pentagon reports of operations targeting drug trafficking vessels in the Caribbean, including strikes since early September 2025. While Tehran Times frames it provocatively, multiple U.S. sources confirm the activity occurred around mid-October near Venezuelan waters."

Reverse image searches indicate that the footage was first uploaded to TikTok that day and we’ve pinpointed the precise coordinates from where it was taken to a military base close to the Camp Pendleton in California, where the anniversary commemoration took place.

The individual behind the account has posted other videos from the same location and appears to be a member of the US military.

The Tehran Times corrected its claim about 24 hours later in a post which has gathered far fewer views than the original false claim, which it still has not deleted.

Grok is no longer repeating the false claim. Asked why the error took place, the chatbot’s creator xAI responded saying “Legacy Media Lies”.