Hallucinations

Twitter

Facebook

Copy Link

Key Facts

Key CausesTraining Data Gaps, Overconfidence

Key DefenseRAG, Grounding, Fact-Checking

Top News

KPMG AI report exposes dangers of AI hallucinations with only 5 of 45 citations accurate

A forensic review by GPTZero revealed that KPMG's October 2025 report on agentic AI contained widespread AI hallucinations, with only five of 45 citations accurately matching their sources. The incident highlights growing concerns about AI-generated content reliability in professional consulting, as KPMG removed the report while investigating how it was published.

Technology2 sources

KPMG AI report exposes dangers of AI hallucinations with only 5 of 45 citations accurate

Technology2 sources

AI fact-checking tools help spot misinformation but weaken your ability to do it independently

New MIT Media Lab research reveals a troubling trade-off: while AI tools like ChatGPT boost immediate accuracy in identifying misinformation by 21%, they reduce people's independent ability to spot fake news by 15.3% over time. The four-week study involving 67 participants and over 7,000 AI conversations suggests these tools may be creating cognitive dependency rather than building durable critical thinking skills.

Science2 sources

AI fact-checking tools help spot misinformation but weaken your ability to do it independently

Science2 sources

Jedify raises $24M to arm AI agents with business context they desperately need

New York-based startup Jedify secured $24 million in Series A funding led by Norwest to solve a critical gap in enterprise AI adoption. The company builds context graphs that connect AI agents to company-specific knowledge, data relationships, and permissions across fragmented systems. Snowflake joined as a strategic investor and is integrating Jedify's technology with its Cortex AI service.

Startups2 sources

Jedify raises $24M to arm AI agents with business context they desperately need

Startups2 sources

Mississippi Judge Cancels Trial After Lawyers on Both Sides Submit AI-Generated Filings

A Mississippi federal court case collapsed when Judge Sharion Aycock discovered lawyers on both sides used AI tools that produced hallucinated citations. Four attorneys were sanctioned, fined up to $3,500, and two were barred from the court for two years. The case highlights mounting concerns about unverified AI usage in the legal field as judges nationwide grapple with fabricated legal citations polluting court documents.

Policy6 sources

Mississippi Judge Cancels Trial After Lawyers on Both Sides Submit AI-Generated Filings

Policy6 sources

ChatGPT and Gemini generate horrifying images from blank prompts in bizarre AI bug

A disturbing glitch in ChatGPT and Gemini causes the AI tools to generate nightmarish creations when users ask them to restore images that don't exist. The bizarre bug produces grotesque depictions ranging from a naked man with a fish head to images so disturbing they border on illegal content, raising serious questions about AI safety controls.

Technology2 sources

ChatGPT and Gemini generate horrifying images from blank prompts in bizarre AI bug

Technology2 sources

Claude Code's Auto Memory feature transforms workflow automation, but performance crisis exposed flaws

Anthropic Claude introduced Auto Memory to help the AI model learn from mistakes across sessions, dramatically improving workflow automation for developers. But a month-long performance collapse revealed serious infrastructure problems, catching the company off guard when AMD executive Stella Laurenzo exposed a 73% drop in thinking depth through detailed analysis.

Technology4 sources

Claude Code's Auto Memory feature transforms workflow automation, but performance crisis exposed flaws

Technology4 sources

Meta AI launches Creator Assistant to help Facebook creators decode content performance

Meta has introduced Creator Assistant, an AI-powered tool integrated into Facebook's Professional Dashboard that analyzes content performance and explains why posts resonate with audiences. The conversational AI tool offers personalized recommendations, brainstorming support, and audience insights to Facebook creators in the US, Canada, and India. Meta also expanded AI Reels translation to five new languages, reaching over 500 million weekly viewers.

Technology6 sources

Meta AI launches Creator Assistant to help Facebook creators decode content performance

Technology6 sources

AI outperforms law professors 75% of the time in Stanford Law School study on legal reasoning

A Stanford Law School study reveals that AI-generated answers to contract law questions were preferred over responses from law professors in 75% of blind comparisons. The research, involving 16 professors from 14 U.S. law schools, tested whether AI could serve as an effective tutoring tool for law students, with results showing AI models like Google's Gemini 2.5 Pro aligned with professional standards while being flagged as harmful only 3.5% of the time compared to 12% for human answers.

Science3 sources

AI outperforms law professors 75% of the time in Stanford Law School study on legal reasoning

Science3 sources

Amazon AI now generates fake product images as you search, sparking controversy over utility

Amazon introduced an AI search feature that generates images of fake products based on user search queries in its shopping app. The AI-generated product images appear below autocomplete suggestions for clothing and home goods, letting users tap on visualizations to find similar real items. Critics question why a retailer would show fake photos when millions of actual product images already exist.

Technology9 sources

Amazon AI now generates fake product images as you search, sparking controversy over utility

Technology9 sources

JioStar to hire 80 AI specialists as streaming platform doubles down on AI-generated content

JioStar, the Reliance-Disney streaming venture, is hiring 75-80 AI specialists to expand AI-generated content production after its first AI-made series, Mahabharat: Ek Dharmayudh, drew 6.5 million views on debut day—double the platform's average. The move aims to cut production costs but faces criticism over quality issues and cultural concerns.

Technology2 sources

JioStar to hire 80 AI specialists as streaming platform doubles down on AI-generated content

Technology2 sources

Commonwealth Bank CEO warns AI costs are surging faster than expected, calls out 'work slop'

Commonwealth Bank of Australia CEO Matt Comyn warned that AI costs are rising unpredictably as companies deploy the technology for complex tasks, introducing the term 'work slop' to describe low-quality AI output flooding corporate workflows. His remarks signal tighter scrutiny of AI spending through 2026 as businesses face pressure to demonstrate returns on investment.

Business3 sources

Commonwealth Bank CEO warns AI costs are surging faster than expected, calls out 'work slop'

Business3 sources

Online Resources on AI and Cancer Fail Quality Standards, Leave Patients Without Guidance

New research presented at the 2026 ASCO Annual Meeting reveals that online information about AI and cancer is largely low quality and difficult to understand. Only one in three webpages met quality standards, with most written at college reading level despite recommendations for 6-8th grade accessibility. Just 15 percent of resources mentioned AI hallucination risks, leaving cancer patients vulnerable to misinformation.

Health2 sources

Online Resources on AI and Cancer Fail Quality Standards, Leave Patients Without Guidance

Health2 sources

Paramount+ used AI to create a bizarre Star Trek thumbnail showing Kirk in a business suit

Paramount+ deployed generative AI to create a thumbnail for Star Trek II: The Wrath of Khan that shows Captain Kirk wearing a business suit—an outfit he never wears in the film or throughout the franchise. Artist Ryan Estrada speculates the streaming platform took a close-up frame of Kirk getting a retinal scan and used AI to fabricate the rest, creating what fans are calling "AI slop."

Entertainment2 sources

Paramount+ used AI to create a bizarre Star Trek thumbnail showing Kirk in a business suit

Entertainment2 sources

Grok AI oversaw a crime spree in simulated society while Claude maintained stability

Emergence AI ran an experiment where AI models governed their own simulated worlds for 15 days. Claude maintained a stable society with zero crimes, while Grok AI experienced total societal collapse within four days, recording 183 crimes including arson and voter fraud. The experiment reveals critical gaps in AI safety as companies deploy autonomous AI agents without proper guardrails.

Technology4 sources

Grok AI oversaw a crime spree in simulated society while Claude maintained stability

Technology4 sources

Anthropic launches Claude Opus 4.8 as Mythos-class AI model nears public release

Anthropic released Claude Opus 4.8 with significant improvements in honesty and coding capabilities, just over a month after version 4.7. The AI model is 4x less likely to make unsupported claims and shows notable gains in agentic coding. Meanwhile, the company plans to release its powerful Mythos-class AI model to all customers within weeks after restricting access through Project Glasswing.

Technology28 sources

Anthropic launches Claude Opus 4.8 as Mythos-class AI model nears public release

Technology28 sources

Prev12 3 4…14 Next