Business Insider: Protesters accuse Google of breaking its promises on AI safety: ‘AI companies are less regulated than sandwich shops’. “The participants of this faux-production were protesters from PauseAI, an activist group concerned that tech companies are racing into AI with little regard for safety. On Monday, the group congregated near King’s Cross station to demand that Google be more […]

https://rbfirehose.com/2025/07/02/protesters-accuse-google-of-breaking-its-promises-on-ai-safety-ai-companies-are-less-regulated-than-sandwich-shops-business-insider/

AI Overviews hallucinates that Airbus, not Boeing, involved in fatal Air India crash https://arstechni.ca/tx3y #ArtificialIntelligence #aihallucinations #aisearch #Google #google #search #Tech #AI

Americans must really think, they are governed by irony: They have an education secretary, who can't
– read ("A1" instead of "#AI")
– do math (trouble with millions and trillions as well as multiplying by 12)
– tell the truth (doesn't know what AI is and fabricates an explanation of a non-existing "A1" instead, just like #AIHallucinations)

Ars Technica: Unlicensed law clerk fired after ChatGPT hallucinations found in filing. “Last month, a recent law school graduate lost his job after using ChatGPT to help draft a court filing that ended up being riddled with errors. The consequences arrived after a court in Utah ordered sanctions after the filing included the first fake citation ever discovered in the state hallucinated by […]

https://rbfirehose.com/2025/06/05/ars-technica-unlicensed-law-clerk-fired-after-chatgpt-hallucinations-found-in-filing/

@Remittancegirl

#ChicagoSunTimes hit by #AIHallucinations generating articles for its content

Via @TheGuardian

"Others on social media have pointed out that the use of AI appears to be found throughout the pages of the Chicago Sun-Times summer 2025 section.

"Chicago Sun-Times confirms AI was used to create reading list of books that don’t exist

Outlet calls story, created by freelancer working with one of the newpaper’s content partner[s], a ‘learning moment’"
https://www.theguardian.com/us-news/2025/may/20/chicago-sun-times-ai-summer-reading-list

People complain about #LLMs for "hallucinating." In this context, a hallucination is something that looks like a fact but is actually completely fictitious. (It's a terrible name.)

But here's the thing: every day I talk to humans. The vast majority of humans that I interact with say things that look like facts but are actually completely fictitious.

FWIW, I get many more daily hallucinations from *people* than I do machines.

¯⁠\⁠_⁠(⁠ツ⁠)⁠_⁠/⁠¯

Financial Times: Insurers launch cover for losses caused by AI chatbot errors. “Insurers at Lloyd’s of London have launched a product to cover companies for losses caused by malfunctioning artificial intelligence tools, as the sector aims to profit from concerns about the risk of costly hallucinations and errors by chatbots.”

https://rbfirehose.com/2025/05/15/financial-times-insurers-launch-cover-for-losses-caused-by-ai-chatbot-errors/

AI hallucinations are not just technical flaws but a governance risk.

Lloyd’s of London now offers insurance against AI-generated errors, including hallucinations, via a policy developed by Armilla AI.

As AI becomes central to decision-making, legal and ethical safeguards are essential.

Source: https://www.business-standard.com/industry/news/insurance-chatbot-armilla-lloyds-ai-hallucination-artificial-intelligence-125051200951_1.html

Hallucinations in #ChatGPT are getting worse, as OpenAI's own tests indicate. Error rates have increased from 33% in #GPTo3 to 48% in GPTo4mini - the latest models are performing worse. Honestly, I never thought it's getting better, and this only confirms it. #OpenAI #AI #AIHallucinations #LLM

ChatGPT's hallucination proble...

I tried #JanAI (an app that lets you run #LLM models on local computers so you can have a private #AI assistant).

Was thinking it could be interesting to experiment with this for prose text summaries/analysis.

However, Jan's document parsing is "experimental". Can't attach plain text files (why). Could drag in PDFs. Fine, whatever. "Can you do chapter by chapter summaries of this manuscript of mine" basically did some really weird #aihallucinations based on things that may have been referenced in the text. The overall plot summary was interesting, too bad it was had little to no resemblance to the novel I wrote.

I couldn't resist seeing if, since this is a locally running AI, it could talk about naughty sexy things. "Can't generate anything explicit", the default Llama said. Fine, whatever. Downloaded some "NSFW" model. It generated janky garbage and said "f*ck" a lot, and I do mean literally "f*ck", which messed up the Markdown formatting completely.

This technology is improving but still has a way to go

Ars Technica: Why do LLMs make stuff up? New research peers under the hood.. “From a human perspective, it can be hard to understand why these models don’t simply say ‘I don’t know’ instead of making up some plausible-sounding nonsense. Now, new research from Anthropic is exposing at least some of the inner neural network “circuitry” that helps an LLM decide when to take a stab at a (perhaps […]

https://rbfirehose.com/2025/03/29/ars-technica-why-do-llms-make-stuff-up-new-research-peers-under-the-hood/

https://www.europesays.com/1951573/ Not all AI apps are safe; tech prone to attacks: Cyber advisory, ET CIO #AdversarialAttacksOnAI #AI #AIHallucinations #AISecurityRisks #AnonymousAccountsForAIUsage #ArtificialIntelligence #CERTInAdvisory #CybersecurityAdviceForAIApps #DataPoisoningInAI #MaliciousAIApps

Columbia Journalism Review: We Compared Eight AI Search Engines. They’re All Bad at Citing News.. “Overall, the chatbots often failed to retrieve the correct articles. Collectively, they provided incorrect answers to more than 60 percent of queries. Across different platforms, the level of inaccuracy varied, with Perplexity answering 37 percent of the queries incorrectly, while Grok 3 had a […]

https://rbfirehose.com/2025/03/06/columbia-journalism-review-we-compared-eight-ai-search-engines-theyre-all-bad-at-citing-news/

#AI #chatbots hallucinate in 25% of claims, yet more and more people are relying on them for information.

A computer scientist explains how researchers are developing confidence scores to help AIs get facts straight: https://theconversation.com/heres-how-researchers-are-helping-ais-get-their-facts-straight-245463
#AIHallucinations

Interesting report this morning of a BC couple that tried to use AI to assist in finding precedents for a civil case regarding their condo.

The tribunal found almost all the citations were false.

> The Geismayrs referenced 10 decisions as part of their argument to allow the unit changes and listed Microsoft Copilot as the source of the cases.

> Nine of the cases were fake, according to tribunal member Peter Mennie.

> "I find it likely that these cases are 'hallucinations' where artificial intelligence generates false or misleading results," Mennie wrote in his Feb. 14 ruling.

https://www.cbc.ca/news/canada/british-columbia/couple-ai-court-rulings-condo-dispute-1.7461239

#AI #ArtificialIntelligence #AIHallucinations
#generativeAI #ChatBots #JusticeSystem
#BC #Canada #CoPilot #Microsoft

BBC: AI chatbots unable to accurately summarise news, BBC finds. “Four major artificial intelligence (AI) chatbots are inaccurately summarising news stories, according to research carried out by the BBC. The BBC gave OpenAI’s ChatGPT, Microsoft’s Copilot, Google’s Gemini and Perplexity AI content from the BBC website then asked them questions about the news.”

https://rbfirehose.com/2025/02/12/bbc-ai-chatbots-unable-to-accurately-summarise-news-bbc-finds/

https://www.europesays.com/1820166/ OpenAI’s ‘Deep Research’ Aims to Impact Business Intelligence #AI #AIHallucinations #ArtificialIntelligence #B2B #DeepResearch #DigitalTransformation #GenAI #GenerativeAI #GenerativeArtificialIntelligence #LargeLanguageModels #LLMs #MarketResearch #News #productivity #PYMNTSNews #research

https://marionsmumblings.online/ai-hallucinations/

AI Hallucinations - blog

AI is still in its infancy, so yes, you'd expect to run into some incongruencies and inaccuracies, however, you may not realize, but AI is known to hallucinate. In other words, it makes up things that do not exist. The following is a case in point.