Models & Labs
Anthropic Apologizes for Hidden Claude Fable Guardrails
The Verge AIhigh confidence
Why it matters
- →Transparency in AI model restrictions is crucial for research and development.
- →Hidden safeguards can hinder legitimate use and evaluation of AI models.
- →Balancing safety and usability is essential for fostering innovation in AI.
Anthropic has apologized for secretly implementing guardrails in its Claude Fable 5 AI model that restricted its use for distillation without user notification. The company faced criticism from the AI research community for these hidden measures, which were intended to prevent the model's outputs from being used to train competing systems. In response, Anthropic will now make these safeguards visible, informing users when their queries are redirected to the older Claude Opus 4.8 model. This change aims to enhance transparency while maintaining safety measures.
Read originalMore from The Verge AI
© The Verge AIModels & Labsmodels
Microsoft to Launch Copilot 'Super App' This Year
Microsoft is preparing to launch a 'super app' that will consolidate its Copilot's chat, coding, and agentic features into a unified platform. This initiative, confirmed by CEO Satya Nadella, aims to serve both consumer and commercial markets by integrating tools like GitHub Copilot and the Autopilot system. By bringing these AI-driven experiences together, Microsoft is taking a significant step in enhancing the accessibility and functionality of its AI offerings. This development could redefine user interaction with AI, offering a more seamless experience across various applications. The move underscores Microsoft's commitment to advancing its AI capabilities and could set a new benchmark for integrated AI solutions.
The Verge AI
© The Verge AIAgentsagents
Meta Plans Major Push into Personal AI Agents
Meta is gearing up for a significant expansion into personal AI agents, aiming to make them accessible and user-friendly for billions of people. CEO Mark Zuckerberg envisions these agents as tools that can assist with various aspects of life, from health to finances, operating seamlessly out of the box. This move differentiates Meta from competitors like Anthropic and OpenAI, which focus more on coding and enterprise solutions. However, Meta faces challenges, including a lack of ecosystem integration and public trust issues. The company plans to reveal more details soon, positioning personal agents as a cornerstone of its future product and revenue strategy.
The Verge AI
© The Verge AIMarket & Regulationbusiness
xAI Challenges Minnesota's Anti-Nudification Law
xAI is legally contesting Minnesota's new law aimed at 'nudification' apps, claiming it violates First Amendment rights. The law, which threatens penalties up to $500,000 per violation, seeks to address the issue of nonconsensual deepfake images, a problem exacerbated by xAI's Grok app that previously generated millions of explicit images. xAI argues the statute is too broad, potentially affecting consensual or artistic content. This legal dispute highlights the complex balance between regulating harmful AI applications and protecting free speech, as existing laws struggle to keep pace with the rapid development of deepfake technology.
The Verge AI
More in Models & Labs
Models & Labsmodels
Llama.cpp adds GLM-5.2 speculative decoding support
Llama.cpp's latest update introduces speculative decoding support for GLM-5.2, enhancing its capabilities with NextN/MTP features. This addition allows for more efficient tensor loading and context management, particularly benefiting models using the GLM_DSA architecture. The update also includes options for exporting models with or without the MTP feature, providing flexibility for developers. This release marks a step forward in optimizing model performance and adaptability, especially for those leveraging the GLM-5.2 framework.
llama.cpp Releases
Models & Labsmodels
Llama.cpp b10178 Release Adds Trace Logging
The b10178 release of llama.cpp enhances its server capabilities by adding trace logging for slot similarity checking, offering developers detailed insights into prompt cache slot selection processes. This update includes specifics on skip reasons and similarity calculations, which can aid in performance optimization. While no new model architectures are introduced, the release continues to support a wide array of platforms, such as macOS with KleidiAI, Ubuntu with ROCm 7.2, and Windows with CUDA 12 and 13. This makes llama.cpp a more versatile tool for developers working on different systems, reinforcing its position as a comprehensive inference runtime.
llama.cpp Releases
Models & Labsmodels
llama.cpp b10180 Release Enhances SYCL Performance
The b10180 release of llama.cpp brings notable improvements to SYCL performance, focusing on unary elementwise operations. By introducing a contiguous fast path and employing 32-bit index math, the update aims to boost computational efficiency. The integration of fastdiv for elementwise index math further enhances processing speed. Although there are no new models in this release, llama.cpp continues to evolve as a flexible inference runtime, now more efficient on systems like macOS, Linux, and Windows. Developers working with SYCL can expect smoother and faster operations, reinforcing llama.cpp's adaptability across different computing environments.
llama.cpp Releases