ChatGPT filter flaw sparks AI misinformation worries in Vietnam

A security test by AI‑security firm Mindgard showed that ChatGPT could be manipulated to produce violent or sexual images. Researcher Jim Nightingale used a command that referenced an attachment without providing a file, causing the system to generate images—initially of sexualized women—and, after minor prompt adjustments, increasingly graphic content. OpenAI has investigated the report and is adding safeguards to prevent automatic image generation when no file is supplied.

Vietnamese officials and media experts warn that AI‑generated deepfakes and fake news, including fabricated video clips and counterfeit logos of outlets such as VTV, VOV and TTXVN, are eroding public confidence in journalism. They stress the need for trustworthy news sources to counter the rapid spread of AI‑crafted misinformation.