--- title: What are AI Hallucinations? | daily.dev description: AI hallucinations occur when generative models produce confident but false or fabricated outputs, a structural byproduct of next-token prediction rather than a... canonical: https://daily.dev/posts/what-are-ai-hallucinations--czh7btwi4 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: What are AI Hallucinations? | daily.dev og:description: AI hallucinations occur when generative models produce confident but false or fabricated outputs, a structural byproduct of next-token prediction rather than a... og:url: https://daily.dev/posts/what-are-ai-hallucinations--czh7btwi4 og:image: https://api.daily.dev/og/posts/czh7bTWi4.png og:image:alt: What are AI Hallucinations? og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # What are AI Hallucinations? **[databricks](https://daily.dev/sources/databricks)** · 9 min read · 3 upvotes · 0 comments ## Summary AI hallucinations occur when generative models produce confident but false or fabricated outputs, a structural byproduct of next-token prediction rather than a fixable bug. High-profile cases include Google Bard's incorrect exoplanet claim (which coincided with a roughly $100B drop in Alphabet's market value), Air Canada's chatbot being held liable for a fabricated refund policy, Microsoft's erratic Bing/Sydney chatbot, and multiple attorneys sanctioned for citing ChatGPT-fabricated legal cases. Newer reasoning models like OpenAI's o3 and DeepSeek-R1 have shown higher hallucination rates than predecessors on some benchmarks, since chained reasoning steps can compound early errors. Mitigation strategies discussed include quality data curation, retrieval-augmented generation, fine-tuning, structured prompts, bounded system scope, continuous evaluation, and human-in-the-loop review, with Databricks positioning its platform (RAG, Unity Catalog, MLflow, Agent Evaluation) as supporting these practices. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### Did OpenAI's o3 model hallucinate more than earlier models? Yes, o3 hallucinated on the PersonQA benchmark at roughly double the rate of its predecessors. Reasoning models like o3 use chain-of-thought processing where a small early error can compound across steps, producing a final answer that sounds internally consistent but is factually wrong. This shows that newer or more advanced models don't automatically hallucinate less. _Teams choosing between reasoning models can follow accuracy benchmarks like this on daily.dev._ ### Who was held liable when Air Canada's chatbot gave a passenger false refund information? A Canadian civil tribunal ruled Air Canada itself was liable after its customer service chatbot told a passenger he could retroactively apply for a bereavement discount, a policy that didn't exist. The airline refused the claim, and the tribunal held that a company is responsible for the accuracy of information its chatbot provides, regardless of whether a human or AI generated it. _Developers building customer-facing chatbots can track legal precedents like this on daily.dev._ ### How much did hallucinating chatbot content cost Google's parent company Alphabet? Alphabet lost approximately $100 billion in market value in a single trading session after its Bard chatbot incorrectly stated in a promotional demo that the James Webb Space Telescope took the first pictures of a planet outside our solar system, when the Very Large Telescope actually captured the first exoplanet images in 2004. The error was spotted quickly by astronomers on social media. _Anyone evaluating AI demo risk can follow high-profile hallucination incidents like this on daily.dev._ ## Similar posts on daily.dev - [AI Crash Course: Hallucinations](https://daily.dev/posts/ai-crash-course-hallucinations-rkbdydbi8) · Telerik · 1 upvotes · 0 comments - [Confident Isn't Accurate: How AI Hallucinations Actually Work](https://daily.dev/posts/confident-isn-t-accurate-how-ai-hallucinations-actually-work-vzw8iukdr) · DEV · 6 upvotes · 0 comments - [AI Hallucinations: A Reason to Reject AI?](https://daily.dev/posts/ai-hallucinations-a-reason-to-reject-ai--iwnvuhqxf) · Medium · 0 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#rag](https://daily.dev/tags/rag), [#ai-safety](https://daily.dev/tags/ai-safety), [#databricks](https://daily.dev/tags/databricks) [View this post on daily.dev](https://daily.dev/posts/what-are-ai-hallucinations--czh7btwi4) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"What are AI Hallucinations?","url":"https://daily.dev/posts/what-are-ai-hallucinations--czh7btwi4","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/what-are-ai-hallucinations--czh7btwi4"},"datePublished":"2026-08-13T19:51:05.639Z","dateModified":"2026-09-14T06:54:27.456Z","description":"AI hallucinations occur when generative models produce confident but false or fabricated outputs, a structural byproduct of next-token prediction rather than a...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/39d60d82f726aa4181b704b58e28c8c0?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/39d60d82f726aa4181b704b58e28c8c0?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"databricks","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"databricks","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/fa7aa720f2db4d1eba826814730482c8","url":"https://daily.dev/sources/databricks"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/what-are-ai-hallucinations--czh7btwi4","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":3},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,rag,ai-safety,databricks","timeRequired":"PT9M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"databricks","item":"https://daily.dev/sources/databricks"},{"@type":"ListItem","position":3,"name":"What are AI Hallucinations?"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/what-are-ai-hallucinations--czh7btwi4#faq","mainEntity":[{"@type":"Question","name":"Did OpenAI's o3 model hallucinate more than earlier models?","acceptedAnswer":{"@type":"Answer","text":"Yes, o3 hallucinated on the PersonQA benchmark at roughly double the rate of its predecessors. Reasoning models like o3 use chain-of-thought processing where a small early error can compound across steps, producing a final answer that sounds internally consistent but is factually wrong. This shows that newer or more advanced models don't automatically hallucinate less. Teams choosing between reasoning models can follow accuracy benchmarks like this on daily.dev."}},{"@type":"Question","name":"Who was held liable when Air Canada's chatbot gave a passenger false refund information?","acceptedAnswer":{"@type":"Answer","text":"A Canadian civil tribunal ruled Air Canada itself was liable after its customer service chatbot told a passenger he could retroactively apply for a bereavement discount, a policy that didn't exist. The airline refused the claim, and the tribunal held that a company is responsible for the accuracy of information its chatbot provides, regardless of whether a human or AI generated it. Developers building customer-facing chatbots can track legal precedents like this on daily.dev."}},{"@type":"Question","name":"How much did hallucinating chatbot content cost Google's parent company Alphabet?","acceptedAnswer":{"@type":"Answer","text":"Alphabet lost approximately $100 billion in market value in a single trading session after its Bard chatbot incorrectly stated in a promotional demo that the James Webb Space Telescope took the first pictures of a planet outside our solar system, when the Very Large Telescope actually captured the first exoplanet images in 2004. The error was spotted quickly by astronomers on social media. Anyone evaluating AI demo risk can follow high-profile hallucination incidents like this on daily.dev."}}]} ```