AI-Powered 'Insights' Feature for the Los Angeles Times Allegedly Justifies Ku Klux Klan’s History

The Los Angeles Times removed its AI-generated “insights” feature after it is alleged to have produced a defense of the Ku Klux Klan. The AI reportedly framed the hate group as a product of societal change rather than an extremist movement. The AI tool, developed by Perplexity and promoted by owner Patrick Soon-Shiong, was designed to provide “different views” on opinion pieces.

Risk classification

• Primary risk domain: 1 Discrimination & Toxicity
• Primary risk subdomain: 1.2 Exposure to toxic content

The AI system exposed readers to toxic content by generating a defense of the Ku Klux Klan, a known white supremacist hate group, minimizing its ideological threat.

Additional risk subdomains

• 7.3 Lack of capability or robustness: The AI system failed to perform reliably, mislabeling political bias, omitting critical context, and citing duplicate or incorrect sources.
• 3.1 False or misleading information: The tool generated misleading summaries and counterperspectives that misinformed readers about the factual context of the articles.

Causal factors

• Entity: AI
• Intent: Unintentional
• Timing: Post-deployment

The generation of pro-KKK defense was an unexpected and unintended outcome of the AI system's goal to provide alternative perspectives after it was deployed to the live website.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Risk Level 3: Limited Risk. The Insights tool is an AI system that generates content (summaries and perspectives) and interacts with users, falling under transparency obligations. The LA Times included a disclaimer stating 'The following content is AI-generated and was not created by the editorial staff' to notify users of the AI's nature.

AI system and alleged parties

• AI system: Graphene, Insights (Perplexity AI)
• AI purpose: Automated Content Curation; Writing Assistant
• Behaviour type: Assistant
• Alleged developer: Perplexity
• Alleged deployer: Patrick Soon-Shiong, Los Angeles Times
• Alleged harmed parties: Los Angeles Times readers, Los Angeles Times, General public

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Minor. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Negligible
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Minor, indirect Negligible
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Negligible, indirect Negligible
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Negligible
• Epistemic: direct Minor, indirect Minor
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Malicious content

Reported: Yes, the report explicitly describes the AI tool generating a defense of the Ku Klux Klan.

Directly caused: The AI-powered Insights tool directly generated a counterperspective framing the 1920s KKK as a product of 'white Protestant culture' responding to societal changes rather than an explicitly hate-driven movement, minimizing its threat.

Inferred additional harm: Readers who viewed the column during the few hours it was live were exposed to this white supremacist apologism, potentially causing offense or distress.

Epistemic

Reported: Yes, the report describes the AI tool generating misleading comments, incorrect bias labels, and faulty citations.

Directly caused: The tool labeled a right-leaning op-ed as centrist, failed to include key context about Donald Trump's actions, and cited mediocre or duplicate sources, sometimes articulating points missing in the cited sources.

Indirectly caused: The union noted that publishing unvetted AI content risks delivering inaccuracies that misinform readers and erode confidence in the news.

Inferred additional harm: Readers who relied on the tool's summaries and bias ratings may have been misinformed about the political orientation and factual context of the opinion pieces.

People affected

• Occurrences reported: 1
• People reportedly exposed: 1000

Potential causes

Management

• Rushed Deployment of AI Tool: Management launched the experimental tool rapidly to court new audiences.
• Owner Unaware of AI Failures: Billionaire owner was unaware of the pro-KKK output hours after launch.

Technology

• AI Accuracy and Reliability Issues: AI chatbots and models remain plagued by accuracy and hallucination issues.
• Flawed Source Citation Logic: The tool cited mediocre sources, duplicated citations, or misled readers.

Data Inputs

• Biased Historical Training Data: Training on decades of historical content can replicate past biases.

Human Factors

• Lack of User Verification: Readers and editors trust AI outputs without checking underlying sources.

Process and Methods

• Insufficient Real-time Oversight: Scaling the tool prevented real-time editorial validation of all responses.
• Inadequate Source Validation: No robust filter to prevent citing low-quality or misleading sources.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide clear, consistent details about the launch of the 'Insights' tool, the specific pro-KKK text it generated, the reaction of the owner and the union, and other factual errors like mislabeling political bias. The role of the Graphene AI system is explicitly stated, and the timeline of events is well-defined.
• Ambiguities identified: The exact number of readers who viewed the pro-KKK text before it was taken down is not specified.
• Alternative interpretations: None. The incident is clearly an AI alignment/robustness failure where a tool designed to show 'opposing views' ended up generating hate group defense.

An AI-powered editorial tool deployed by the Los Angeles Times generated a defense of the Ku Klux Klan and produced factual inaccuracies before being removed. The incident represents an unintentional alignment and robustness failure of a commercial AI system with negligible direct national security impact, though it highlights ongoing concerns regarding AI-generated misinformation and public trust.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: United States
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Long-term. The feature was quickly removed, and the incident represents an ongoing, long-term strategic concern regarding AI deployment in journalism rather than an active crisis.
• Autonomy: Human-supervised. The AI system generated content autonomously, but operated within a human-controlled publishing environment where editorial staff could intervene and remove the feature.
• Novelty: Established threat. AI models generating biased, inaccurate, or toxic content when attempting to summarize or provide alternative viewpoints is a well-established technical failure mode.

Impact by dimension

• Physical security: Negligible. No physical threat, kinetic attacks, or critical infrastructure impact occurred as a result of this incident.
• Information security: Negligible. The incident was an internal, unintentional editorial tool failure rather than a coordinated foreign information warfare operation or intelligence breach.
• Sovereignty: Negligible. No government operations, electoral systems, border controls, or constitutional processes were impacted by this media-specific tool failure.
• Economic security: Negligible. No threat to financial systems, critical supply chains, or national technological competitive advantage was identified.
• Societal stability: Minor. The AI tool generated a defense of a prominent hate group (the KKK), which touches on societal cohesion, but the content was removed within hours, resulting in minimal lasting societal impact.