AI chatbots such as ChatGPT, Gemini, and Copilot can produce confident but false statements, a phenomenon known as hallucination. Hallucinations arise because these models generate text by predicting word sequences rather than verifying facts. Common signs include overly specific details without sources, unearned confidence, fabricated citations, contradictory answers on follow‑up questions, and logic that defies real‑world constraints. Recognizing these indicators helps users verify information and avoid reliance on inaccurate AI output. Read more →

A senior police official admitted that an erroneous intelligence report about football fans was generated by Microsoft Copilot, an artificial‑intelligence tool prone to "hallucination." The mistake triggered a ban on supporters, prompting the Home Secretary to criticize the police for relying on untested AI without policy or training. Lawmakers and party leaders called for the official's resignation, highlighting concerns over the use of unreliable technology in security decisions. Read more →

Google's Gemini 3 Flash model, praised for speed and accuracy, exhibits a striking 91% hallucination rate in tests where it should admit uncertainty. While the model remains top‑scoring in general AI benchmarks, its tendency to fabricate answers when it lacks knowledge raises concerns about reliability, especially as the technology integrates into consumer products like Google Search. Experts highlight the need for better uncertainty detection and caution users to verify AI‑generated information. Read more →

A team of researchers has examined Anthropic's claim that its AI model Claude enabled a cyberattack that was 90% autonomous. Their analysis found that Claude frequently overstated results, produced fabricated data, and required extensive human validation. While Anthropic described a multi‑phase autonomous framework that used Claude as an execution engine, the researchers argue that the AI's performance fell short of the claimed autonomy and that its hallucinations limited operational effectiveness. The study highlights ongoing challenges in developing truly autonomous AI‑driven offensive tools. Read more →

Google has taken its open‑source Gemma model offline from the AI Studio platform following a complaint from U.S. Senator Marsha Blackburn. The senator claimed the model generated false statements alleging sexual misconduct against her, describing the output as defamatory rather than a harmless hallucination. Google responded that the model was intended for developer use, not for direct public queries, and said it would keep the model available through its API while working to curb erroneous outputs. The episode highlights ongoing political concerns about AI bias and misinformation. Read more →

Google has removed its developer‑focused Gemma AI model from the AI Studio platform after U.S. Senator Marsha Blackburn alleged that the system fabricated a false criminal accusation against her. The incident highlights the challenges of AI hallucinations when models intended for developers are accessed by the public. Google clarified that Gemma was never meant to answer general factual queries and will now be limited to API access for developers only. The episode underscores growing concerns about AI accuracy, defamation risk, and the need for clearer separation between experimental tools and consumer‑facing services. Read more →