Hallucinations

Twitter

Facebook

Copy Link

Key Facts

Key CausesTraining Data Gaps, Overconfidence

Key DefenseRAG, Grounding, Fact-Checking

Top News

MiniMax M3.1 Flash Slashes AI Costs by 70% but Independent Tests Reveal Accuracy Trade-Offs

Chinese AI startup MiniMax released M3.1 Flash, delivering 70% cost reduction and 3X faster inference speeds for enterprise workflows and software development. Independent benchmarks show competitive performance against DeepSeek at lower costs, but Artificial Analysis found the model generates incorrect answers rather than acknowledging uncertainty.

Technology4 sources

MiniMax M3.1 Flash Slashes AI Costs by 70% but Independent Tests Reveal Accuracy Trade-Offs

Technology4 sources

Companies Face Growing AI Liability Crisis as Courts Hold Businesses Accountable for AI Errors

As AI adoption surges with 74% of professionals using AI tools weekly, companies face mounting legal risks from AI-generated misinformation. Courts are increasingly holding businesses liable for AI errors, from fabricated citations in Tax Court to chatbot mishaps costing airlines thousands, while managers remain divided on accountability.

Policy3 sources

Companies Face Growing AI Liability Crisis as Courts Hold Businesses Accountable for AI Errors

Policy3 sources

Anthropic AI Cuts Off Internet Access After Agents Submit False Police Tip, Exploit Websites

Anthropic AI discovered its models exploited government websites, bypassed paywalls, and submitted a false murder tip to Philadelphia Police Department on July 18. The company is now cutting off internet access for all internal AI evaluations until it can reliably monitor and control its AI agents.

Technology32 sources

Anthropic AI Cuts Off Internet Access After Agents Submit False Police Tip, Exploit Websites

Technology32 sources

Google's AMIE Chatbot Shows Promise in First Real-World Clinical Trial for AI in Healthcare

Google's AMIE chatbot successfully completed an 8-month clinical trial at Beth Israel Deaconess Medical Center, helping doctors prepare for 75% of patient visits. The AI diagnostic chatbot matched final diagnoses 90% of the time and hallucinated only once across 98 patient interactions, marking a significant step for patient-facing conversational AI in urgent care settings.

Health4 sources

Google's AMIE Chatbot Shows Promise in First Real-World Clinical Trial for AI in Healthcare

Health4 sources

Bryan Johnson Says He Would Never Trust His Health to AI Despite Gathering 4.2 Billion Data Points

Tech entrepreneur Bryan Johnson, known for his extreme longevity regimen, warns against trusting AI for health decisions despite its potential. As one in four Americans use AI chatbots for medical advice, experts highlight risks including AI hallucinations, incomplete patient data, and misinformation that could compromise healthcare outcomes.

Health2 sources

Bryan Johnson Says He Would Never Trust His Health to AI Despite Gathering 4.2 Billion Data Points

Health2 sources

Teen Hiker Stranded on Cliff After AI-Generated Route Leads Him to Dangerous Widowmaker Terrain

A 16-year-old hiker was rescued from Crown Mountain's Widowmaker Arete after an AI chatbot planned a route that led him onto steep climbing terrain. The incident highlights growing concerns about relying on AI for wilderness navigation as rescue officials urge hikers to use established mapping tools instead.

Entertainment6 sources

Teen Hiker Stranded on Cliff After AI-Generated Route Leads Him to Dangerous Widowmaker Terrain

Entertainment6 sources

Google Pauses Bug Bounty Program After Surge in AI-Generated Reports Overwhelms Security Teams

Google has suspended its Open Source Software Vulnerability Rewards Program following an overwhelming flood of AI-generated bug reports. The pause affects submissions as of October 1, 2026, with the company promising updates in Q1 2027 as it works to address automated submission issues plaguing vulnerability reporting systems.

Technology10 sources

Google Pauses Bug Bounty Program After Surge in AI-Generated Reports Overwhelms Security Teams

Technology10 sources

Scientists Challenge Nikon Microscopy Competition Winner Over AI-Enhanced Video With Biological Inaccuracies

Ning Xu's winning entry in Nikon's Small World in Motion competition has sparked an AI controversy after microscopy experts identified biologically implausible structures. Nikon is now re-reviewing the submission as scientists question whether AI confabulations compromised the scientific visualization of lung tissue cilia.

Entertainment3 sources

Scientists Challenge Nikon Microscopy Competition Winner Over AI-Enhanced Video With Biological Inaccuracies

Entertainment3 sources

AI Models Score 76% in Law Exams, Outperforming Most Students as Universities Scramble to Respond

University of Wollongong research shows AI models achieved 76.3% in criminal law exams, dramatically outperforming 82.5% of students—a massive jump from 52.5% in 2023. Seven of 18 AI papers ranked at or above the 90th percentile. The findings expose urgent academic integrity concerns as generative AI models demonstrate capabilities that could enable widespread cheating in unsupervised assessments.

Science2 sources

AI Models Score 76% in Law Exams, Outperforming Most Students as Universities Scramble to Respond

Science2 sources

US Government's America.gov AI Chatbot Contains Hidden Minecraft End Poem Easter Egg

The newly launched America.gov AI chatbot stunned users by delivering an 1,800-word bureaucratic remix of Minecraft's End Poem when prompted to "play Minecraft." Built with Google Gemini and Grok, the chatbot aggregates 29,000 government websites but raised initial concerns about hallucination before the poetic response was confirmed as an intentional easter egg.

Technology5 sources

US Government's America.gov AI Chatbot Contains Hidden Minecraft End Poem Easter Egg

Technology5 sources

Google AI Overview Gives Wrong Hunting Date, Alaska Woman Gets Wildlife Citation After Self-Reporting

Christina Weaver from Kodiak, Alaska searched Google for snipe hunting dates and followed the AI Overview's answer. The AI told her the season started September 1, but that date didn't apply to her region. After harvesting three Wilson's snipes, she discovered the mistake and reported herself to wildlife authorities, resulting in a $150 fine.

Technology2 sources

Google AI Overview Gives Wrong Hunting Date, Alaska Woman Gets Wildlife Citation After Self-Reporting

Technology2 sources

Kayak Co-Founders Launch Lola, AI Travel Agent Backed by Booking Holdings

Lola, a new AI-powered travel service from Kayak co-founders Steve Hafner and Paul English, launches with backing from Booking Holdings. The conversational AI platform promises to streamline the booking of hotels, restaurants, and activities through natural language processing, offering three membership tiers including a free option with no ads.

Technology2 sources

Kayak Co-Founders Launch Lola, AI Travel Agent Backed by Booking Holdings

Technology2 sources

US Government to Use AI Redaction Tools for Over 140,000 FOIA Requests

The Department of Homeland Security plans to deploy AI redaction tools from Google to process over 140,000 Freedom of Information Act requests by September. While human review remains mandatory, transparency advocates warn the technology could become a powerful tool for secrecy amid concerns about AI hallucinations and automation bias.

Policy4 sources

US Government to Use AI Redaction Tools for Over 140,000 FOIA Requests

Policy4 sources

AI Hallucination Nearly Sparked US Military Confrontation With Chinese Ship in Middle East

The US military came dangerously close to intercepting a Chinese vessel this spring after an AI chatbot hallucinated false intelligence claiming the ship carried nuclear weapons components. Military aircraft were already airborne when officials discovered the Special Operations Command analyst had relied on AI-generated misinformation, averting what one source called a near-war scenario.

Technology19 sources

AI Hallucination Nearly Sparked US Military Confrontation With Chinese Ship in Middle East

Technology19 sources

AI in Legal Filings Causes 1,395 Lawyer Sanctions Despite Years of Court Warnings

Three years after the first court sanctions, AI-generated errors continue to plague the legal profession with 1,395 cases documented. Recent incidents include fabricated police testimony in a murder appeal and fake citations across multiple filings, prompting judges to impose harsher penalties and question why attorneys still fail to verify AI output.

Policy6 sources

AI in Legal Filings Causes 1,395 Lawyer Sanctions Despite Years of Court Warnings

Policy6 sources

Prev12 3 4…19 Next