Anthropic's watermark, Twitch data policy, and Apple's China AI model

Anthropic has detailed its new invisible text watermarking system, which uses a secret key and preceding words to embed statistical patterns, making it difficult to remove. The company also revealed that two unreleased successors to Claude Mythos 5, named Model 1 and Model 2, are being used internally, with Model 2 showing increased capability. Separately, Twitch faced backlash for its default opt-in policy for using livestream content for AI training, requiring users to actively disable it. In other news, Apple has reportedly trained its own large language model for the Chinese market with Alibaba's assistance, marking a strategic shift to navigate local regulations and control its data pipeline. AI

IMPACT Developments in AI watermarking, data usage policies, and regional model training highlight evolving industry practices and regulatory challenges.

RANK_REASON The item discusses multiple AI-related developments from different companies without focusing on a single originating event.

Read on dev.to — LLM tag →

• Alibaba
• Anthropic
• Claude
• Claude Mythos 5
• DeepSeek
• EU AI Act
• Google DeepMind
• Model 1
• Model 2
• Qwen
• SynthID-Text
• Twitch

• model release
• policy
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →