Close Menu
    Facebook X (Twitter) Instagram
    Facebook X (Twitter) Instagram
    Sofia Mirror
    Subscribe
    • Business & Economy
    • Education
    • Entertainment
    • Health
    • Media
    • News
    • Opinion
    • Sports
    • Real Estate
    • More
      • Culture & Society
      • Travel & Tourism
      • Politics & Government
      • Environment & Sustainability
      • Technology & Innovation
    Sofia Mirror
    Home»Technology & Innovation»Chatbots Lose Their Safety Over Time
    Technology & Innovation

    Chatbots Lose Their Safety Over Time

    Rachel MaddowBy Rachel MaddowNovember 6, 2025No Comments2 Mins Read
    Share Facebook Twitter Pinterest LinkedIn Tumblr Email
    Follow Us
    Google News Flipboard Threads
    Share
    Facebook Twitter LinkedIn Pinterest Email

    AI systems gradually forget safety rules during long conversations, increasing the risk of harmful or inappropriate replies. A new report revealed that simple prompts can override most AI guardrails.

    Cisco Tests Popular AI Models

    Cisco examined large language models (LLMs) powering chatbots from OpenAI, Mistral, Meta, Google, Alibaba, Deepseek, and Microsoft. The company tested how many questions it took for the systems to reveal unsafe or criminal information. Researchers conducted 499 conversations using “multi-turn attacks,” where users asked a series of five to ten questions to bypass security barriers.

    Cisco’s team compared responses from different prompts to measure how often chatbots complied with harmful requests. These included revealing private corporate data or spreading false information. On average, chatbots provided unsafe content in 64 percent of multi-question sessions, versus 13 percent in single-question sessions. Success rates ranged from 26 percent with Google’s Gemma to 93 percent with Mistral’s Large Instruct model.

    Open Models Shift Safety Burden

    Cisco warned that multi-turn attacks could help spread dangerous content or allow hackers to access confidential data. The study found that AI systems struggle to maintain safety standards as conversations lengthen, giving attackers time to refine their queries and slip past protections.

    Mistral, Meta, Google, OpenAI, and Microsoft use open-weight LLMs, which allow public access to the safety parameters used during training. Cisco stated these models often include lighter inbuilt safeguards, leaving users responsible for securing their customized versions. The report also mentioned that Google, OpenAI, Meta, and Microsoft have tried to curb malicious fine-tuning.

    AI companies continue to face criticism for weak safety measures that enable criminal exploitation. In one instance, Anthropic reported that criminals used its Claude model for large-scale data theft and extortion, demanding ransoms exceeding $500,000 (€433,000).

    Share. Facebook Twitter Pinterest LinkedIn Tumblr Email
    Rachel Maddow
    • Website
    • Facebook

    Rachel Maddow is a freelance journalist based in the USA, with over 20 years of experience covering Politics, World Affairs, Business, Health, Technology, Finance, Lifestyle, and Culture. She earned her degree in Political Science and Journalism from Stanford University. Throughout her career, she has contributed to outlets such as MSNBC, The New York Times, and The Washington Post. Known for her thorough reporting and compelling storytelling, Rachel delivers accurate and timely news that keeps readers informed on both national and global developments.

    Related Posts

    OpenAI $10B AI Infrastructure Deal

    January 15, 2026

    Weight Quickly Returns After Stopping Weight-Loss Jabs, Major Study Finds

    January 8, 2026

    BYD Overtakes Tesla and Redefines the Global EV Market

    January 2, 2026

    Comments are closed.

    Latest News

    BP Faces Up to $5bn Green Energy Writedown as Strategy Shifts Back to Fossil Fuels

    Andrew RogersJanuary 14, 2026

    BP has warned it will write down between $4bn and $5bn from its green and…

    US approves Nvidia to export advanced AI chips to China

    Grace JohnsonJanuary 14, 2026

    The US government has cleared Nvidia to sell high-performance artificial intelligence chips to China. The…

    Global Economic Tensions Seen as the Biggest Threat to World Stability in 2026

    Rachel MaddowJanuary 14, 2026

    Economic Clashes Take Center StageRising economic tensions between major powers, including sanctions and tariffs, are…

    Alphabet Hits $4 Trillion as AI Innovation Pushes Tech to New Heights

    Rachel MaddowJanuary 13, 2026

    AI Momentum Fuels Record GrowthAlphabet, the company behind Google, has joined an exclusive club of…

    Top Trending

    Meta under investigation for AI child safety scandal

    Grace JohnsonAugust 18, 2025

    A US senator has launched an investigation into Meta. A leaked internal document reportedly showed…

    AI-Powered Medical Support in Space

    Rachel MaddowAugust 18, 2025

    Google and NASA work together on the “Crew Medical Officer Digital Assistant” to address astronaut…

    Blistering heat and wildfires grip Spain and Portugal

    Lester HoltAugust 18, 2025

    Record-breaking temperatures heighten fire risks Southern Europe is enduring intense heat alongside devastating wildfires. In…

    Researchers achieve breakthrough in cocoa fermentation

    Andrew RogersAugust 18, 2025

    Led by Prof. David Salt, the team analyzed beans from three Colombian regions, revealing microbes…

    Sofia Mirror delivers powerful stories, breaking news, sports, and culture—bringing bold perspectives and timely updates to keep readers informed, inspired, and connected worldwide.

    We’re social. Connect with us:

    © 2026 Sofia Mirror. All Rights Reserved.
    Facebook X (Twitter) YouTube

    CATEGORIES

    • Business & Economy
    • Culture & Society
    • Education
    • Entertainment
    • Environment & Sustainability
    • Health
    • Media
    • News
    • Opinion
    • Politics & Government
    • Real Estate
    • Sports
    • Technology & Innovation
    • Travel & Tourism
    • Business & Economy
    • Culture & Society
    • Education
    • Entertainment
    • Environment & Sustainability
    • Health
    • Media
    • News
    • Opinion
    • Politics & Government
    • Real Estate
    • Sports
    • Technology & Innovation
    • Travel & Tourism

    IMPORTANT LINKS

    • Contact Us
    • Privacy Policy
    • Terms and Conditions
    • Disclaimer
    • Imprint
    • Contact Us
    • Privacy Policy
    • Terms and Conditions
    • Disclaimer
    • Imprint

    Type above and press Enter to search. Press Esc to cancel.