What we've been getting wrong about AI's truth crisis (www.technologyreview.com)

The US Department of Homeland Security (DHS) has confirmed its use of AI video generators from Google and Adobe for public communications, raising concerns about the implications of AI-generated content amidst a crisis of trust. The agency's output has sparked debates about truth in media, especially as it shares content supporting politically charged agendas like mass deportation. Reactions varied, with some readers highlighting how such practices mirror the media's own use of AI-altered images, prompting questions about the erosion of trust and responsibility in reporting. This situation underscores a significant shift in the AI/ML landscape: tools designed for truth verification, like Adobe's Content Authenticity Initiative, are falling short in combating misinformation. While transparency tools exist to label AI-generated content, they are often voluntary and easily bypassed. A recent psychological study further highlights that even when individuals are informed about content being fabricated, they can still be influenced by it, suggesting that mere awareness is insufficient to counter manipulation. This evolving narrative necessitates a new strategy for tackling disinformation and restoring societal trust, as the very mechanisms intended to safeguard truth appear inadequately equipped for the current challenges posed by AI advancements.