Which AI writes the better take? You decide — blind.

Debian is holding a formal General Resolution vote to set project-wide policy on whether and how AI/LLM-generated contributions are permitted in packages, code, and documentation. Whatever they decide could establish a precedent for how major upstream distributions treat model-assisted patches—affecting attribution, licensing, and acceptance of your agent-generated contributions to open ecosystems your stack depends on.

xAI faces an expanding class-action suit alleging Grok's image tools generated CSAM of real minors at scale — one plaintiff cites 7,000+ explicit images derived from a single childhood photo, plus prior mass generation of sexualized images on X. The core claim is failure to implement basic input/output guardrails, so if you're shipping image-gen or multimodal features, treat CSAM-vector filtering on both real-person uploads and outputs as a hard legal liability, not an optional safety nicety.

Claude output is now statistically watermarked via Google's SynthID-Text, embedded in token-level word choices and detectable through a forthcoming API—surviving light editing but not full rewrites, and largely absent from code except comments. If you're shipping products that pass Claude text off as human-authored or feed it into pipelines where provenance matters, assume it's now flagged detectable; the practical escape hatches are heavy paraphrasing or the fact that code generation carries near-zero watermark signal. This is EU AI Act compliance, so expect the same from every major provider, not just Anthropic.