Hallucinations

Twitter

Facebook

Copy Link

Key Facts

Key CausesTraining Data Gaps, Overconfidence

Key DefenseRAG, Grounding, Fact-Checking

Top News

AI chatbots deliver inaccurate election advice, raising concerns about electoral interference

Research during Hungary's parliamentary elections reveals AI chatbots like ChatGPT and Gemini provided inaccurate and unreliable voting advice to users. The systems recommended parties not running, omitted relevant parties, and showed significant political bias. In 90% of cases, ChatGPT failed to recommend the opposition Tisza party to aligned voters, while consistently recognizing Fidesz-aligned profiles, exposing serious regulatory gaps in AI oversight.

Policy2 sources

AI chatbots deliver inaccurate election advice, raising concerns about electoral interference

Policy2 sources

ChatGPT and AI chatbots are quietly rewiring how the human brain makes decisions

Recent studies from Carnegie Mellon University and Microsoft Research reveal a troubling pattern: when people consult ChatGPT for answers, they stop admitting uncertainty and grow dangerously confident in incorrect responses. The research exposes how AI tricks our brains through automation bias, making confident errors feel like verified facts.

Entertainment2 sources

ChatGPT and AI chatbots are quietly rewiring how the human brain makes decisions

Entertainment2 sources

European Parliament builds EPGenAI Hub as AI slop infiltrates legislative work

The European Parliament is launching EPGenAI Hub, an internal AI platform for MEPs and staff, after discovering that roughly 2,100 people use public chatbots daily to draft legislation. The platform will offer models from OpenAI, Meta, Anthropic, and Mistral to prevent sensitive parliamentary material from being fed into uncontrolled public tools. The rollout could begin as early as September 2025.

Technology2 sources

European Parliament builds EPGenAI Hub as AI slop infiltrates legislative work

Technology2 sources

AI advice makes people three times less accurate but twice as confident, new study reveals

Researchers from French and Italian universities discovered that AI advice suppresses people's willingness to say 'I don't know' from 44% to just 3%. Accuracy plummeted from 27% to 9%, while confidence paradoxically surged from 30% to 76%. The study highlights how AI might be harming our critical thinking skills, with experts particularly concerned about children who are growing up with these systems before developing foundational critical thinking skills.

Science3 sources

AI advice makes people three times less accurate but twice as confident, new study reveals

Science3 sources

Finance teams embrace AI but 93% question the insights it generates, revealing deep trust crisis

New research reveals a stark contradiction in AI in finance: while adoption accelerates and over 60% of teams now use real-time operational data, 93% of finance professionals doubt the integrity of AI-generated insights. Concerns span hallucinations, inaccuracies, and bias. Meanwhile, 85% of Indian finance leaders face pressure to demonstrate ROI without clear AI governance frameworks in place.

Business3 sources

Finance teams embrace AI but 93% question the insights it generates, revealing deep trust crisis

Business3 sources

AI Lawsuits Reveal Insurance Gaps as Autonomous Systems Outpace Conventional Coverage

A report by the Artificial Intelligence Underwriting Company reveals that over 90% of insurers' exposure to AI risks sits in silent coverage embedded in conventional insurance policies. Rising AI lawsuits—from Google's $110 million defamation case to Air Canada's chatbot mishap—show how autonomous AI systems are creating liabilities that existing policies were never designed to cover.

Business2 sources

AI Lawsuits Reveal Insurance Gaps as Autonomous Systems Outpace Conventional Coverage

Business2 sources

ZoomInfo Launches GTM Bench, First AI Benchmark Designed for Real-World Sales and Marketing Work

ZoomInfo unveiled GTM Bench, a new benchmark system that evaluates AI agents and large language models on actual go-to-market tasks. Unlike traditional AI benchmarks that test reasoning in controlled environments, GTM Bench scores systems on two critical measures: how much work they complete and whether outputs are grounded in verifiable data. In initial testing, ZoomInfo's GTM.AI achieved a GTM Bench Index of 77, significantly outperforming competitors like Apollo and Exa.

Technology2 sources

ZoomInfo Launches GTM Bench, First AI Benchmark Designed for Real-World Sales and Marketing Work

Technology2 sources

Coinbase AI sends false breaking news alert about match that hadn't started yet

Coinbase's AI-generated breaking news alert told millions that Norway beat Brazil 3-2 in a FIFA World Cup match before kickoff. The AI hallucination exposed critical flaws in how AI systems handle uncertainty in real-time applications, particularly concerning for prediction markets where bettors rely on accurate information. While Norway eventually won 2-1, the fabricated alert highlights ongoing struggles with AI-generated misinformation despite billions spent on development.

Technology2 sources

Coinbase AI sends false breaking news alert about match that hadn't started yet

Technology2 sources

OpenAI seeks investment banking expert to train ChatGPT on Wall Street tasks

OpenAI posted a job listing for a Subject Matter Expert in Investment Banking, signaling its ambition to train ChatGPT for complex financial transactions including mergers, acquisitions, and valuations. The role, offering up to $205K plus equity, aims to define quality standards for AI-assisted banking work and improve model performance on high-stakes financial tasks.

Business2 sources

OpenAI seeks investment banking expert to train ChatGPT on Wall Street tasks

Business2 sources

HalluSquatting attack exploits AI hallucinations to build botnets by tricking coding assistants

Security researchers from Tel Aviv University revealed HalluSquatting, an AI security threat that exploits AI hallucinations to trick AI agents into executing malicious code. The attack targets popular AI coding assistants like Cursor, GitHub Copilot, and Gemini CLI, with hallucination rates reaching 85% for repositories and 100% for trending skills. Attackers register fake resource names predicted by large language models, enabling AI-enabled botnets at unprecedented scale.

Technology5 sources

HalluSquatting attack exploits AI hallucinations to build botnets by tricking coding assistants

Technology5 sources

ChemGraph AI framework automates chemistry simulations to accelerate battery and materials research

Researchers at Argonne National Laboratory developed ChemGraph, an open-source AI framework that automates computational chemistry workflows using large language models and agent-based automation. Built on the Aurora exascale supercomputer, the framework allows scientists to describe research problems in plain language while AI agents handle complex simulation tasks, potentially accelerating battery development, combustion research, and critical material extraction.

Technology3 sources

ChemGraph AI framework automates chemistry simulations to accelerate battery and materials research

Technology3 sources

AI chatbots deliver confident financial advice that's often wrong, leaving investors burned

Millions are turning to AI chatbots for personal finance guidance, but the results are troubling. A 2025 survey reveals 19% of Americans lost over $100 following AI financial advice, with Gen Z hit hardest at 27%. Finance experts warn the real danger isn't obvious errors—it's that AI chatbots sound so authoritative that people never seek professional help, missing critical details that surface only years later.

Science9 sources

AI chatbots deliver confident financial advice that's often wrong, leaving investors burned

Science9 sources

Tencent Hy3 beats larger rivals at half the size with Apache 2.0 license, reshaping open AI

Tencent released Hy3, a 295-billion-parameter Mixture-of-Experts AI model under Apache 2.0 license, removing geographic restrictions that plagued earlier Chinese models. With hallucinations cut from 12.5% to 5.4%, the model leads in agentic search and tool orchestration while trailing GLM-5.2 in coding. The license shift signals a new era for enterprise AI deployments.

Technology2 sources

Tencent Hy3 beats larger rivals at half the size with Apache 2.0 license, reshaping open AI

Technology2 sources

Dartmouth Study Reveals AI in Healthcare Can Increase Physician Workload With Flawed Responses

A large-scale Dartmouth study analyzing 146,000 patient-doctor conversations found that AI in healthcare often creates more work than it saves. AI-generated responses frequently contain errors and irrelevant details, forcing physicians to spend more time editing AI mistakes than writing messages themselves. However, researchers developed a technique that improves accuracy by 33% and reduces editing by 26%.

Health2 sources

Dartmouth Study Reveals AI in Healthcare Can Increase Physician Workload With Flawed Responses

Health2 sources

AI Companies Are Hiring Philosophy Graduates to Solve Their Biggest Problems

Philosophy graduates are finding unexpected career paths in the AI industry, where their training in consciousness, ethics, and reasoning helps tackle critical challenges. From AI safety and alignment to machine consciousness and AI hallucinations, philosophers are applying centuries-old analytical methods to cutting-edge technology problems at companies like Google DeepMind and OpenAI.

Technology2 sources

AI Companies Are Hiring Philosophy Graduates to Solve Their Biggest Problems

Technology2 sources

Prev12 3 4…16 Next