Skip to content
Submit incident
AII-1002-00883
ReportedReported by fewer than three independent publishers, with no official finding.

Grok chatbot provides inaccurate and contradictory information on Iran-Israel conflict

A study by the Atlantic Council found that Grok, xAI's chatbot integrated into the X platform, provided inaccurate and contradictory responses about the 12-day Iran-Israel conflict, including inconsistent claims about destroyed airport images and misidentification of airport locations. The research analyzed 130,000 multilingual messages and found the chatbot had difficulties authenticating AI-generated content.

25 June 2025 · xAI · Reliability and hallucination
25 June 2025
AII-1002-008832 sources ↓
Quick facts
Category
Reliability and hallucination
Country
Israel
Source
T13 (teletrece)
Date
25 June 2025
Drafted by
Raidu Incident Detection

What happened

A study by the Atlantic Council's digital research laboratory analyzed 130,000 messages in multiple languages on the X platform and found that Grok 'had difficulties authenticating content generated by AI' during the Iran-Israel conflict.

When responding to questions about AI-generated images of a destroyed airport that accumulated millions of views on X, Grok oscillated within the same minute between denying the airport's destruction and confirming it had been hit. In some responses Grok claimed damage was from a missile fired from Yemen; in others it falsely identified the airport as being in Beirut, Gaza, or Tehran.

Reported impact

The study found that Grok's responses 'during the first days of the conflict between Israel and Iran revealed significant failures and demonstrated the chatbot's limitations in providing accurate, reliable and coherent information in times of crisis.'

With the gradual disappearance of journalist fact-checking on platforms like Facebook, increasing numbers of internet users rely on AI tools like Grok to find reliable information, but the responses obtained sometimes misinform them.

Index analysis

According to experts, Grok has previously made errors in fact-checking about the recent India-Pakistan conflict and during demonstrations in Los Angeles against immigration raids ordered by US President Donald Trump.

Last month, the AI service also caused controversy by incorrectly referring to a 'white genocide' in South Africa, an error that xAI later attributed to 'unauthorized modification'.

Classification

Category
Reliability and hallucination
Industry
Media, social platforms and marketing
Country
Israel
Harm status
Potential harm
Intent
Unintentional
Timing
Post-deployment
Harm types
Public interest
Harmed
General public
Failure modes
Hallucination, Misidentification, Agent misaction
Consequences
None recorded
AI systems
Grok

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

T13 (teletrece)
Estudio revela "fallos significativos" del chatbot Grok (de Elon Musk) en respuestas a guerra Irán-Israel
25 June 2025
Zócalo Saltillo
Estudio revela fallos de Grok, IA de Elon Musk, en respuestas a guerra Irán-Israe | Periódico Zócalo | Noticias de Saltillo, Torreón, Piedras Negras, Monclova, Acuña
25 June 2025