To shield kids, California hikes fake nude fines to $250K max (arstechnica.com)

California has signed two laws (effective Jan 1, 2026) aimed at protecting children from harms tied to AI companion bots and deepfake pornography. Companion‑bot platforms—explicitly calling out services like ChatGPT, Grok and Character.AI—must publish protocols for identifying and addressing users’ suicidal ideation or self‑harm, report statistics to the state Department of Public Health on how often they issued crisis‑prevention notifications (and publish those stats on their sites), prohibit bots from claiming to be therapists, and implement child‑safety measures such as break reminders and blocking sexually explicit images for minors. Separately, penalties for knowingly distributing nonconsensual AI‑generated sexually explicit material have been raised so victims (including minors) can seek up to $250,000 per deepfake from third parties. For the AI/ML community, the rules signal stronger regulatory expectations around detection, logging, transparency and liability: platforms will need robust conversational intent detection (suicidality/self‑harm classifiers), reliable image moderation and age‑gating, secure audit trails for reported interventions, and possibly watermarking or provenance tools to combat deepfakes. The higher statutory damages also raise distributional risk for third‑party hosts and aggregators, which could drive faster adoption of technical countermeasures (watermarks, forensic detectors, provenance metadata) and force tradeoffs between safety, privacy, and product design—potentially becoming a blueprint for other jurisdictions.