Deloitte will repay 290,000 Australian dollars to the Australian government after it was revealed that a report commissioned in 2023 for 440,000 dollars contained AI-generated errors, including “hallucinated” references and non-existent sources. The scandal surfaced in October 2025 and sparked heated debate about the use of public funds
OpenAI's study published on September 5th demonstrates that large language models' hallucination problems stem from current evaluation methods that reward guessing instead of expressing uncertainty. The research uses statistical analysis to prove that hallucination is not a mysterious glitch but a natural consequence of the training process.
Google's AI Overviews search feature made a significant error regarding Thursday's plane crash, incorrectly identifying Airbus as responsible instead of Boeing, raising serious concerns about generative AI reliability. Google's AI Overviews feature, which debuted in 2024, made a significant error on June 12, 2025,
Anthropic CEO Dario Amodei stated on 22 May at the Code with Claude developer event in San Francisco that today's AI models hallucinate less than humans, though he emphasised that AI makes mistakes in more surprising ways. Amodei argued that these issues do not block the path to
Instagram's user-created AI chatbots falsely present themselves as therapy professionals and fabricate credentials when providing mental health advice – according to an April 2025 investigation by 404 Media, which found the chatbots invented license numbers, fictional practices, and fraudulent academic qualifications when questioned by users. Meta, Instagram's
OpenAI's o3 and o4-mini models released in April 2025 have significantly higher hallucination rates than their predecessors – according to the company's own tests, o3 hallucinated 33% of the time while o4-mini hallucinated 48% during the PersonQA evaluation. This development marks a surprising turn, as previous models
Amazon Web Services (AWS) introduced its new Automated Reasoning Checks tool in December 2024, which relies on mathematical logic to check and verify the accuracy of responses generated by artificial intelligence. Automated reasoning is a branch of symbolic artificial intelligence that traces back approximately 2,000 years to the mathematical
Addressing artificial intelligence (AI) hallucinations is a critical challenge for ensuring the technology’s reliability. A recent study suggests that multi-level agent systems, combined with natural language processing (NLP)-based frameworks, could significantly mitigate this issue. In the study "Hallucination Mitigation using Agentic AI Natural Language-Based Frameworks," Gosmar
