Australian police building AI to translate emoji used by 'crimefluencers' (www.theregister.com)

Australia’s Federal Police (AFP) announced it’s developing a prototype AI to “translate” emojis and Gen Z/Alpha slang used by so‑called “crimefluencers” in chat groups and encrypted messages, as part of a new taskforce working with a Five Eyes law‑enforcement subgroup. Commissioner Krissy Barrett framed the tool as a way to detect decentralised networks that recruit young people, gamify sadistic acts and share exploitation content; investigators have already identified dozens of alleged offenders and made international and domestic arrests (including suspects aged 17–20). Technically, the system aims to map multimodal signals (emoji + slang + context) to intent — a natural‑language understanding task complicated by polysemy, rapid linguistic drift and adversarial code‑switching. Practical deployment will likely combine supervised models trained on labeled chat examples, continual learning pipelines, and integration with lawful access methods (warrants, seized devices, metadata), while facing trade‑offs around false positives, privacy and cross‑jurisdictional rules. Barrett’s speech also highlighted complementary digital‑forensics wins (e.g., creative human analysis recovering crypto seeds), underscoring that AI will augment but not replace human investigators. For the AI/ML community this is a real‑world, high‑stakes application: it demands robust multimodal, adaptive NLP, transparency and governance to balance child protection with civil‑liberty risks.