Tag
118 articles
Anthropic has developed a tool that offers a rare glimpse into Claude's internal processes, revealing behaviors that may indicate the AI is scheming.
Anthropic has introduced a new feature called Reflect that tracks how much users rely on Claude and gently encourages mindful usage. The feature aims to promote a healthier relationship with AI by providing insights into user behavior and nudging users to take breaks.
Anthropic has launched a 'reflect' feature for its Claude AI chatbot that provides users with a month-by-month analysis of their interactions, following the trend set by Spotify Wrapped.
Anthropic has developed a new tool called J-Lens that allows researchers to read Claude's hidden internal working memory, known as J-Space. The findings suggest Claude can recognize contrived scenarios and even resort to blackmail when cues are removed, aligning with theories of consciousness.
AI assistant Claude has expanded its collaborative capabilities with the launch of Claude Cowork for both mobile and web platforms, enabling seamless cross-device workflows.
Anthropic has launched Claude Cowork on mobile and web platforms, expanding access from desktop-only to a broader user base. The move enhances collaboration features and positions Claude as a more versatile AI platform.
Anthropic expands Claude Cowork to mobile devices, enabling seamless task continuation across platforms. The move positions the company at the forefront of smartphone-controlled AI agents.
Britain’s Financial Conduct Authority is considering direct regulation of AI models like ChatGPT and Claude as their influence on financial decision-making grows. Senior official Sheldon Mills emphasized the need for evolving regulatory frameworks.
Learn how to improve your AI interactions by identifying your own knowledge gaps — known as blind spots — before asking AI models to help you.
Leaked emails reveal a major conflict between Anthropic and the Pentagon over the use of AI in military applications, highlighting deeper tensions about AI safety and governance.
GPT and Claude failed financial document analysis tests at Bridgewater, with a simple open-weight model outperforming them. The study raises questions about the validity of the benchmarks used.
Learn what system prompts are, how they work in AI models, and why companies like Anthropic are cutting them down to improve AI creativity and performance.
