This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
Following President Biden's withdrawal from the 2024 presidential race, the AI chatbot Grok on X generated false information claiming that Vice President Kamala Harris had missed ballot deadlines in nine U.S. states. This misinformation was widely shared on social media, prompting five secretaries of state to send an open letter to Elon Musk demanding immediate corrections and the implementation of reliable election information sources. The false information persisted for over a week before being corrected.
After President Joe Biden stepped aside as a presidential candidate on July 21, 2024, the AI chatbot Grok on X reportedly falsely informed users that Vice President Kamala Harris missed the ballot deadline in nine states. This misinformation, which spread widely on social media, prompted secretaries of state from five U.S. states to urge Elon Musk to address the problem.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI systems that inadvertently generate or spread incorrect or deceptive information, which can lead to inaccurate beliefs in users and undermine their autonomy. Humans that make decisions based on false beliefs can experience physical, emotional or material harms
Also classified as
The chatbot failed to perform reliably on factual queries, hallucinating incorrect ballot deadlines due to a lack of capability in verifying real-time political facts.
Grok generated and spread false information regarding election ballot deadlines, leading to inaccurate beliefs and public confusion about candidate eligibility.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The incident was caused by the Grok AI system generating incorrect information post-deployment, which was an unexpected and unintended outcome of its operation.
Added categories peak at 4: Severe (Epistemic Harm, Direct)
The Grok AI chatbot generated and spread false information claiming Vice President Kamala Harris missed ballot deadlines in nine states. This misinformation reached millions on social media, requiring intervention from five Secretaries of State. This incident highlights the national security risks of AI-driven election misinformation and its potential to disrupt democratic processes and public trust.
Grok is described as an 'AI search assistant' and chatbot that provides information and answers questions for users, acting to augment human capabilities rather than taking independent action.
Limited Risk: The system is a chatbot and generates synthetic content, which falls under Risk Level 3 requiring transparency obligations so users know they are interacting with an AI.
Grok is explicitly described as an AI chatbot and search assistant designed to answer user queries and interact with users.
The reports provide clear, consistent, and detailed accounts of the false information generated by Grok, the specific states mentioned, the duration of the issue, and the response from public officials. There is little ambiguity regarding the AI's role or the nature of the misinformation.
None
Technology
Unfiltered Chatbot Design
Grok was built to answer 'spicy' queries without standard safety filters.
Inadequate Hallucination Safeguards
The model generated false claims about ballot deadlines without verification.
Data Inputs
Lack of Nonpartisan Data Feed
The system did not integrate with trusted databases like CanIVote.org.
Outdated Training Data
The model relied on outdated rules that did not reflect current deadlines.
Human Factors
Rapid User Sharing
Premium subscribers quickly shared false AI outputs, reaching millions.
Process and Methods
Delayed Correction Workflow
The company took over a week to correct the chatbot's false output.
Inadequate Output Validation
There was no robust process to flag and block false election information.
Management
Prioritizing Engagement Over Safety
Management designed Grok to be unfiltered, sidelining safety safeguards.
Reduced Moderation Resources
Staff reductions at X weakened the ability to monitor and correct AI errors.
Dismissive Response Strategy
Leadership responded slowly and dismissively to election official warnings.
Regulatory Environment
Lack of Election AI Standards
No regulatory mandates enforced accuracy or safe boundaries for election AI.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
