Australia rolls out deep‑fake detection training as AI predicts World Cup semifinals

The University of New South Wales in Australia has introduced a training program that improves participants’ ability to spot AI‑generated faces by about 40 percent. Lead researcher Amy Dawel said the program uses a series of real and synthetic photographs and teaches six visual cues, such as symmetry, to distinguish deepfakes. She warned that as detection improves, AI generation will also evolve, making the “cat‑and‑mouse” game harder for the public.

Separately, consultancy Hubler ran an experiment using large‑language models – GPT‑4, Claude Sonnet, and Gemini 2.5 Pro – to forecast the FIFA 2026 World Cup semifinals. The models correctly identified three of the four finalists (France, Argentina and Spain) but differed on the fourth, with Gemini incorrectly favoring Brazil. Hubler’s CEO Angélica Guerrero highlighted how model choice and data governance can change outcomes, noting implications for business forecasting and risk analysis.

Both pieces illustrate the growing influence of artificial‑intelligence tools in security, media literacy, and predictive analytics.