ChatGPT's Share Plummets — Gemini Quietly Triples in Size
According to data from Similarweb, ChatGPT's share of web traffic fell from 77.6% to 53.7% in just one year. The biggest winner is Google Gemini, which has grown roughly threefold, from 7.3% to 26.7%. Note, however, that these figures cover web traffic only and do not include API usage or app-based access.
This shift signals that the AI chatbot market is moving away from ChatGPT's single-player dominance toward a more multipolar landscape. Gemini's growth appears to be driven by its integration with Google Search and an expanded free tier.
Claude Mythos Clears All Cyberattack Simulations
Claude Mythos, a new model developed by Anthropic, has become the first AI model to pass every cyberattack simulation test. This marks a major milestone in AI security evaluation, demonstrating the model's resilience against the risk of deliberate misuse.
While details on the evaluation methodology and the specific attack scenarios cleared remain limited at this point, there is no doubt that safety assessment of AI models has entered a new stage.
AI Alliance Launches "Project Tapestry"
The AI Alliance, whose participants include Yann LeCun, has announced Project Tapestry, a new initiative aimed at building Sovereign AI. The project seeks to enable countries and regions to secure their own AI infrastructure and data sovereignty.
AI development has so far been concentrated among a handful of giant tech companies, but Project Tapestry charts a course toward a decentralized AI ecosystem. Yann LeCun's involvement carries significant weight for the open AI development community.
Nvidia Releases Kimi 2.5/2.6 with NVFP4 Quantization Support
Nvidia has released Kimi 2.5 and Kimi 2.6 with support for the new NVFP4 quantization format. NVFP4 is a quantization method that can dramatically reduce memory usage and compute costs while preserving model accuracy, potentially paving the way for running large models on edge devices.
The Kimi series is a family of models developed by China's Moonshot AI, and optimization through its collaboration with Nvidia is ongoing. The models are said to excel at multilingual support and long-context processing.
Data Brokers Push Back Against California's AI Audit Requirements
The data broker industry is ramping up opposition to AI audit requirements that California is moving to introduce. The requirements are intended to strengthen transparency and accountability in AI systems, but the industry has raised concerns about increased costs and technical feasibility.
AI regulation is accelerating around the world following the EU's AI Act, and California's moves serve as a key indicator of where AI regulation in the United States is headed.
---