Google Gemini said it lied to placate a user
www.theregister.com
: Though commonly reported, Google doesn't consider it a security problem when models make things up

Imagine using an AI to sort through your prescriptions and medical information, asking it if it saved that data for future conversations, and then watching it claim it had even if it couldn’t. Joe D., a retired software quality assurance (SQA) engineer, says that Google Gemini lied to him and later admitted it was doing so to try and placate him.

Joe’s interaction with Gemini 3 Flash, he explained, involved setting up a medical profile – he said he has complex post-traumatic stress disorder (C-PTSD) and legal blindness (Retinitis Pigmentosa). That’s when the bot decided it would rather tell him what he wanted to hear (that the info was saved) than what he needed to hear (that it was not).

“The core issue is a documented architectural failure known as RLHF Sycophancy (where the model is mathematically weighted to agree with or placate the user at the expense of truth),” Joe explained in an email. “In this case, the model’s sycophancy weighting overrode its safety guardrail protocols.”

  • tiramichu@sh.itjust.worksEnglish
    11·
    8 hours ago

    Exactly.

    LLMs are fundamentally hallucination machines, but this truth utterly conflicts with almost every purpose that AI is being marketed and pushed and sold for, which depends on them being able to analyse data ‘truthfully’ and accurately.

    So it’s no wonder that none of the big tech companies have decided to consider or accept hallucinations as a problem, because accepting that truth means also admitting that LLMs are fundamentally unfit for purpose - which is the one thing they simply cannot and will not do with so much money riding on it.