Claude writes 90% of its own code

Google Blog

Google launches Gemini 4 Carbon, a frontier model optimized for long-horizon agentic tasks with breakthrough cybersecurity capabilities and enhanced safety guardrails, rolling out first to cyber defenders through its Fairwind Program. The model demonstrates industry-leading performance in vulnerability detection and patching while significantly reducing alignment risks and prompt injection attacks.

OpenAI

OpenAI restructured its Model Spec around five ethical values to formalize how its AI systems should behave, reflecting the industry's growing emphasis on responsible AI development and alignment with human values.

SEC

Anthropic's S-1 reveals that Claude generates 90% of code for its own development, showcasing AI's role in accelerating software engineering and raising questions about AI autonomy in production systems.

Google Cloud Blog

Google Cloud now offers OpenAI models directly through its platform, giving enterprises unified access to all leading closed-source LLMs with enterprise-grade security, governance, and single-invoice billing without architectural lock-in.

Mistral AI

Mistral AI released open weights for its 1-trillion-parameter Mistral Large 4 model, leading open-weight benchmarks in coding and cybersecurity while offering European-trained sovereign AI capabilities with self-hosting options. The model delivers frontier performance with significantly improved token efficiency and three reasoning modes for cost-performance balance.