Grok spread misinformation about the shooting in Australia - ForkLog: cryptocurrencies, AI, singularity, the future

# Grok spread disinformation about the shooting in Australia

The AI-based chatbot Grok repeatedly spread disinformation about a mass shooting on Bondi Beach in Australia.

This appears to be an old viral video of a man climbing a palm tree in a parking lot, possibly to trim it, resulting in a branch falling and damaging a parked car. Searches across sources show no verified location, date, or injuries. It may be staged; authenticity is uncertain.

— Grok (@grok) December 14, 2025

On December 14, in Sydney, the annual Hanukkah by the Sea celebration took place — the start of Hanukkah, organized by Chabad of Bondi. About 1000 people participated.

In the second half of the day, two attackers — a father and son in black clothing — opened fire on a crowd at the playground in Archer Park, firing about 50 shots. Police classified this as a terrorist act with an antisemitic motive, the second deadliest mass shooting in Australia’s history.

Sixteen people were killed, and 42 injured.

One of the attackers was disarmed by a bystander — 43-year-old Ahmed al-Ahmad. The incident video circulated on social media, many praised the heroism of the individual.

The xAI chatbot by Elon Musk provided users with false information about the incident. In response to a question about a video showing al-Ahmad fighting the shooter, the AI replied:

“It appears to be an old viral clip of a man climbing a palm tree in a parking lot, possibly to trim it. The branch then fell onto a damaged car. Searches across various sources have not confirmed details about the location, date, or injuries. It may be staged; authenticity is unverified.”

Grok also claimed that a photo showing al-Ahmad was taken on October 7, 2023. It allegedly was held hostage by Hamas for over 700 days and was released in October 2025.

In response to another user question, the chatbot wrote a non-relevant paragraph about whether the Israeli army deliberately targeted civilians in Gaza.

Reports from doctors in Gaza, such as Dr. Nick Maynard in September 2025, describe cases of Palestinian boys and men with gunshot wounds to the groin, suggesting deliberate targeting by Israeli forces. However, Israeli officials, including Netanyahu, have denied similar…

— Grok (@grok) December 14, 2025

In one case, Grok described a video showing a shootout between an attacker and police in Sydney as a clip about Cyclone Alfred.

Among other things, the AI confuses information about the beach incident and the shooting at Brown University, which occurred a few hours before the Australian attack.

The malfunction spread not only to the shooting. Throughout December 14, Grok:

• misidentified well-known football players;
• provided information about acetaminophen use during pregnancy, although asked about pills for its termination;
• discussed the likelihood of Kamala Harris running again in the US presidential election when asked to verify a completely different statement regarding British law enforcement initiatives.

Alternate Reality

Grok has repeatedly been involved in scandals due to incorrect and false statements about various events and matters.

In July, users noticed that the neural network relied on Elon Musk’s opinions when forming responses. This included topics like the Israel-Palestine conflict, abortions, and immigration laws.

Observations suggest that the chatbot was deliberately configured to consider Musk’s political views when answering controversial questions.

Earlier, the billionaire stated that his startup would rewrite “all human knowledge” to train a new version of Grok, since today “there is too much garbage in any baseline model trained on unverified data.”

Subsequently, Grokipedia was created — an AI-based online encyclopedia “focused on truth.”

Recall that in November, users noticed bias in Grok 4.1 — the new model significantly overestimated Elon Musk’s abilities.