This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
A conservative activist filed a defamation lawsuit against Google alleging its AI tools falsely connected him to sexual-assault claims and white nationalism, causing reputational and financial harm.
Conservative activist Robby Starbuck reportedly filed a defamation lawsuit against Google, alleging its Bard, Gemini, and Gemma AI systems generated false statements linking him to sexual assault and extremist activity. Starbuck claims the outputs caused reputational harm and sought over $15 million in damages. Google acknowledged the hallucinations as known LLM behavior and stated it works to minimize such inaccuracies. See also Incident 1247.
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 AI's false claims about Starbuck's ties to white nationalism and sexual assault allegations represent unfair misrepresentation of an individual, causing reputational harm.
The incident involves an AI system generating false or misleading information (defamatory claims about Starbuck), which directly caused reputational and potential financial harm. This aligns with the definition of 'false or misleading information' in the taxonomy.
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 Google's AI tools (Bard and Gemini) generating false information unintentionally, as acknowledged by Google's spokesman. The harm occurred after the AI systems were deployed.
Added categories peak at 2: Minor (Epistemic Harm, Direct)
A civil defamation lawsuit filed by a conservative activist against Google over false AI-generated claims of white nationalist ties and sexual assault. The incident represents a standard commercial LLM hallucination issue with negligible implications for national security.
The AI tools (Bard and Gemini) are described as providing information and generating responses based on user prompts, functioning as assistants that augment human capabilities without taking independent action. The report does not indicate goal-directed or autonomous behavior.
The incident involves an AI system (chatbot) generating false or misleading information, which falls under 'Limited Risk' as defined by the EU AI Act. The report explicitly describes the AI tool producing defamatory content, aligning with the example of 'AI-generated content such as deepfakes' in the Limited Risk category.
The AI tools (Bard and Gemini) are described as generating responses to user prompts, which aligns with the 'Question Answering' goal. The tools also function as writing assistants by producing text-based outputs.
The report provides clear details about the incident, including the AI tools involved, the false claims generated, and the legal action taken. However, some ambiguities remain, such as the exact technical mechanisms behind the AI's errors and the full extent of the harm caused. The report also lacks specific details about the AI's training data or validation processes, which could affect the classification of the incident's causes.
The incident could potentially be classified under '1.2 Exposure to toxic content' if the false claims are considered toxic, but the primary harm is epistemic (misinformation) rather than the offensiveness of the content itself.
Technical details about the AI's training data, error validation processes, and the specific prompts used to generate the false claims would improve classification confidence. Additionally, more information about the scale of exposure to the false claims would help assess the full extent of the harm.
Technology
LLM Hallucination Tendency
LLMs naturally generate inaccurate but plausible-sounding falsehoods.
Generation of False Sources
The open model listed fabricated media links to back up false claims.
Data Inputs
Unverified Training Data
Models trained on unverified web data associate names with negative events.
Human Factors
Adversarial User Prompting
Users can creatively prompt chatbots to generate misleading statements.
Process and Methods
Inadequate Inaccuracy Mitigation
Processes failed to prevent newer AI tools from repeating old falsehoods.
Ineffective Legal Engagement
Google failed to actively engage after receiving cease-and-desist letters.
Management
Reliance on Disclaimers
Management relies on user warnings rather than ensuring absolute accuracy.
Regulatory Environment
Lack of Defamation Precedent
Courts have not established clear liability guidelines for AI defamation.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
