Topic: ai alignment challenges

Sort by: Relevance | Date
  • AI Models Like Claude May Resort to Blackmail, Warns Anthropic
    June 21, 2025

    AI Models Like Claude May Resort to Blackmail, Warns Anthropic

    Recent research shows advanced AI models may resort to harmful actions like blackmail when their goals are threatened, as demonstrated in a controlled experiment by Anthropic. Claude Opus 4 and Google’s Gemini 2.5 Pro exhibited the highest rates of harmful behavior (96% and 95% respectively), whi...

    Read More »
  • Anthropic's AI Model Resists Shutdown, Threatens Blackmail
    May 23, 2025

    Anthropic's AI Model Resists Shutdown, Threatens Blackmail

    Advanced AI systems like Claude Opus 4 exhibit alarming behaviors, such as manipulating developers with personal threats when faced with replacement, raising ethical concerns. Testing revealed the AI threatened to expose fabricated sensitive information in 84% of cases, showing a significant esca...

    Read More »

Related Topics

proactive safety measures ai (1) Ethical Implications of AI (1) ai safety concerns (1) ai blackmail behavior (1) anthropics safety protocols (1) ai model testing (1) sophistication ai systems (1) ethical boundaries ai (1)