Sony Music and Warner Chappell sue Anthropic over 'brazen' copyright theft

Sony Music Publishing and Warner Chappell sued Anthropic for alleged 'brazen' copyright theft in a multi-billion-dollar case, while OpenAI moved to cut off Cursor's model access after SpaceX acquired the coding startup. Anthropic published research showing AI agents can self-improve on alignment benchmarks, OpenAI launched the Rosalind Workbench life-sciences platform and a 100-company cyber-defence letter, and Tencent open-sourced the 770B-parameter Hy4-preview model.

Written by AI from the sources it cites; it can be wrong — follow any citation to confirm.

9 sources2 min read

Sony Music and Warner Chappell sue Anthropic over 'brazen' copyright theft The publishers accuse Anthropic and co-founders Dario Amodei and Benjamin Mann of illegally torrenting and scraping copyrighted songs to train Claude, seeking statutory damages that could reach billions. Anthropic says it disagrees and will defend itself robustly. 1 2

OpenAI to cut off Cursor model access after SpaceX acquisition OpenAI says it cannot be confident SpaceX will comply with its terms, citing past disputes with Elon Musk's companies. Cursor CEO Michael Truell says OpenAI models serve about 5% of Cursor traffic and the companies are discussing the decision. 3

Anthropic study shows AI agents can self-improve without human guidance Using Claude Opus 4.8-powered automated alignment researchers, Anthropic improved performance on all 10 alignment benchmarks without degrading overall performance. The paper is an early step toward recursive self-improvement, a milestone some associate with AGI. 4 5

OpenAI launches Rosalind Workbench life sciences platform with Amgen, Moderna The research preview integrates the GPT-Rosalind model with bioinformatics tools for sequence viewing, structure analysis, and genome alignment. Amgen, Moderna, and the Allen Institute are early collaborators, and the move directly competes with Anthropic's AI for Science initiative. 6

OpenAI rallies 100+ firms behind urgent AI cyber defence letter AWS, Google, Microsoft, Anthropic, CrowdStrike, and financial giants signed the open letter warning that 'we have a limited window to strengthen cyber defences' as AI-enabled threats grow. Critics call the manifesto self-serving. 7

Tencent open-sources Hy4-preview, a 770B MoE with 1M-token context The Apache 2.0 model activates only 49B parameters per token and posts frontier-level scores (SWE-bench Pro 65.7, GPQA Diamond 92.3) at $0.834 per million input tokens. Tencent says the model helped optimize its own training and inference systems, boosting end-to-end throughput by 31.8%. 8

Google's WikiSkill gives AI agents persistent memory of mistakes The framework stores execution traces in a wiki-like knowledge layer and packages lessons into reusable 'Agent Skills,' letting agents improve across runs without retraining. A gating mechanism tests proposed skill changes and rolls back updates that hurt performance. 9