When AI Gets It Wrong: The Microsoft Brandon Hunter Obituary Incident
How a headline calling a deceased NBA player 'useless at 42' exposed critical failures in AI content systems and what designers must learn to prevent similar failures.
In September 2023, Microsoft made a mistake that would reverberate through the tech and media industries. Their MSN news portal published an AI-generated obituary for former NBA player Brandon Hunter that called him "useless at 42" -- a grotesque combination of automated content generation and zero human oversight. This incident serves as a critical case study in why human review remains essential when AI touches sensitive content, and what lessons designers, developers, and content creators must learn from this failure.
For organizations deploying AI in content workflows, the Brandon Hunter incident highlights the urgent need for robust human-in-the-loop systems. Without proper oversight mechanisms, even the most sophisticated AI can produce content that damages brand reputation and causes real harm to affected individuals and their families.
The Incident: What Happened
On September 14, 2023, MSN published an obituary for Brandon Hunter, a former NBA player who had suddenly died at age 42. The AI-generated article contained multiple egregious errors that demonstrated the dangers of automated content without human oversight. The article was part of Microsoft's automated content system, which had been generating news articles and obituaries with minimal human intervention.
The incident quickly went viral on social media, with users expressing shock and outrage at the insensitive headline. The backlash highlighted how automated content systems, when deployed without adequate safeguards, can produce results that no human editor would ever approve.
The Specific Failures
The most shocking error was the headline itself: "Brandon Hunter useless at 42" -- a phrase that appeared to be a direct output of the AI system with no human editing or review. As Business Insider reported, the headline was algorithmically generated based on patterns in the data, with no consideration for the emotional impact of the wording.
Beyond the headline, the article reportedly contained bizarre phrasing and factual errors. Search Engine Land documented how the AI had generated nonsensical statements about Hunter's career and life, mixing information inappropriately and failing to show the basic respect expected in memorial content. The article was swiftly removed after public backlash, but the damage to Microsoft's reputation was already done.
Why This Matters for Content Design
The Brandon Hunter incident is not an isolated case -- it represents a systemic risk when AI systems are deployed for content generation without proper safeguards. Understanding what went wrong helps designers create better systems that prevent similar failures. This case demonstrates the critical gap between AI's ability to generate fluent text and its ability to understand context, tone, and the human impact of content decisions.
For UX professionals, this incident underscores the importance of designing interfaces that prioritize human judgment in sensitive contexts. Our user experience design services emphasize the critical role of human oversight in AI-assisted content workflows.
Why AI Struggles with Sensitive Content
AI language models work by predicting the most likely next word or phrase based on patterns in their training data. This approach works well for many tasks but fails dramatically when dealing with specific factual information about real people. The hallucination problem makes AI particularly dangerous for obituaries, where accuracy about someone's life, career, and family relationships is essential.
As The Atlantic explored in their investigation of AI in the funeral industry, these systems lack the emotional intelligence and contextual understanding that humans bring to sensitive content creation.
The Hallucination Problem
As Josh McQueen of Passare explained, "When we first started testing this, ChatGPT would just make up stories." The AI might assert that someone named Billy was often called "Skippy" and then fabricate an anecdote to explain the fake nickname. This hallucination tendency makes AI particularly dangerous for obituaries, where accuracy about someone's life, career, and family relationships is essential.
A single fabricated detail can cause real harm to grieving families and damage the trust that communities place in memorial content. The Brandon Hunter case demonstrates how AI systems can confidently generate false information without any indication that something is wrong.
Context and Tone Are Hard for AI
Obituaries occupy a unique position in writing -- they must be factual while also being compassionate, celebratory while remaining dignified. The tone must balance respect for the deceased with acknowledgment of the grief felt by survivors. These nuanced social and emotional requirements are difficult for AI systems to understand, let alone replicate.
According to Poynter's analysis, the AI that generated the Hunter obituary apparently failed to recognize that calling someone "useless" -- even if technically accurate in some contexts -- would be profoundly offensive in an obituary. This failure demonstrates the gap between linguistic competence and social awareness that remains in current AI systems.
AI vs Human Content Decision-Making
Understanding the fundamental differences between AI and human content processing helps designers build better systems that leverage the strengths of both while mitigating their respective weaknesses.
AI Decision Process
1. Analyze input data about person 2. Generate content based on patterns 3. Detect keywords (death, obituary) 4. Output content ← NO REVIEW STEP
Human-in-the-Loop Process
1. Analyze input data about person 2. Generate draft content 3. Human reviews for accuracy, tone, appropriateness 4. Human approves or revises 5. Final content published
The Human Element in Memorial Content
As Ian Bogost discovered when writing his own mother's obituary, "A person should not pretend to be a friend, and a computer should not pretend to be a person." The problem isn't just accuracy -- it's authenticity. Obituaries serve as the final public statement about someone's life. They carry emotional weight for families and communities.
When AI generates content that lacks genuine human understanding, it produces something that feels hollow, even when factually correct. This is why our approach to AI-powered content solutions always emphasizes human oversight as a core component of the design.
When AI Can Help Appropriately
This doesn't mean AI has no place in obituary writing. The Atlantic found that when properly configured and reviewed by humans, AI obituary generators can actually produce good results. The key is using AI as a tool for assistance rather than replacement.
Passare's obituary tool requires users to input specific facts about the deceased and then generates a draft that family members can review and revise. The system was adjusted to prevent hallucination by carefully crafting the prompts fed to ChatGPT behind the scenes. The appropriate use case is AI as a starting point that accelerates the writing process while leaving final approval to humans who understand the emotional context.
The Human-AI Collaboration Model
Effective human-AI collaboration for sensitive content follows a structured workflow that ensures human oversight at critical decision points. This model treats AI as an accelerant rather than a replacement for human judgment.
INPUT FACTS → AI GENERATION → HUMAN REVIEW → PUBLISH
The model: Input facts into AI, let it generate a draft, require human review before any publication. Human can revise or reject the content. Only after final human approval should content be published.
Best Practices for AI in Sensitive Content
The Brandon Hunter incident makes clear that any AI system generating content about real people -- especially in sensitive contexts like obituaries, news, or medical information -- must include mandatory human review before publication. Poynter's media ethics analysis emphasizes that this review cannot be optional or easily bypassed; it must be built into the workflow as a required step.
Essential Human Oversight
Organizations deploying AI for content generation should establish clear policies about which types of content require human review, who is responsible for that review, and what criteria must be met before publication. For memorial content specifically, the reviewer should ideally have some understanding of the deceased or access to someone who can verify accuracy.
This oversight requirement should be technically enforced in the content management system, not left to individual discretion. The goal is to create friction that ensures no sensitive content goes live without proper review. This approach aligns with broader content quality and SEO principles that prioritize accuracy and user trust.
Context-Aware Prompts and Guardrails
As Passare discovered, the prompts fed to AI systems can dramatically affect output quality. Organizations should invest time in developing prompts that explicitly prevent the types of errors seen in the Hunter incident. The Atlantic reported on how careful prompt engineering can reduce hallucination and improve tone appropriateness.
This includes prompts that explicitly forbid offensive language, require verification of facts, and maintain appropriate tone. Beyond prompts, technical guardrails can prevent publication of content that triggers specific keywords or patterns associated with problematic outputs.
Quality Control Metrics
Organizations should establish clear metrics for evaluating AI-generated content before publication. The following checklist provides a framework for reviewing sensitive AI-generated content.
Factual accuracy verified
All claims about the subject have been confirmed through reliable sources
No offensive language
Content contains no words or phrases that could cause harm
Tone appropriate
Content matches the expected tone for the context
Key details confirmed
Birth/death dates, family info, career details are accurate
No hallucinations
No fabricated stories, quotes, or events included
Human approval complete
A qualified reviewer has approved the content
Audience appropriate
Content is suitable for the intended audience
Standards met
Content follows organizational guidelines
Lessons for UX and Interface Design
The technical architecture of AI content systems should enforce human review as a required step. This means designing interfaces that don't allow content to be published until a reviewer has taken specific action. The path to publication should literally not exist without human intervention.
For organizations building AI-powered content systems, our custom software development services can help implement these safeguards effectively.
Making Human Review Mandatory
Design interfaces that enforce human review as a required step. This might mean designing interfaces that don't allow content to be published until a reviewer has taken specific action, such as clicking an approval button or checking specific boxes. The path to publication should literally not exist without human intervention.
Key design considerations include: requiring explicit approval actions, showing clear status indicators throughout the review process, and preventing any bypass mechanisms that could allow content to skip review.
Clear Indication of AI Generation
Interfaces should clearly indicate when content has been generated or assisted by AI, both for reviewers and potentially for end users. This transparency helps reviewers understand they need to be especially careful and helps organizations maintain accountability when problems arise.
Visual indicators, audit logs, and metadata tagging all contribute to a culture of transparency around AI-assisted content creation.
Interface Flow for AI-Assisted Content
1. USER INPUTS INFORMATION
Clear form fields for required facts
2. AI GENERATES DRAFT
Status indicator shows 'AI-generated'
3. REQUIRED HUMAN REVIEW
Reviewer sees AI-generated badge, checklist of verification items, cannot proceed without approval
4. REVISION IF NEEDED
Easy editing tools for reviewer, ability to regenerate or modify
5. APPROVAL AND PUBLICATION
Final approval step required, audit trail of who approved
Conclusion: Human-Centered AI
The Microsoft Brandon Hunter obituary incident represents a cautionary tale about deploying AI without adequate human oversight. The technology's ability to generate fluent text does not equate to its ability to generate appropriate, accurate, or respectful content in sensitive contexts.
For designers, developers, and content professionals, the lesson is clear: AI should augment human capabilities rather than replace human judgment. When AI touches content about real people -- particularly content as sensitive as obituaries -- human review must be mandatory, thoughtful, and equipped with the tools needed to catch errors before publication.
The funeral industry's experience with AI obituary tools offers a more balanced perspective. When properly designed with human review built into the workflow, these tools can genuinely help families navigate difficult times while maintaining the dignity and accuracy that memorial content requires. The goal should not be fully automated content generation, but rather human-AI collaboration that leverages the strengths of both.
The "useless at 42" headline will long serve as a reminder of what happens when organizations forget that some content requires human hearts, not just human minds, to get right.