Deception using artificial intelligence is increasingly pervading our online lives. Known cases include altered images of protesters or videos that spread misinformation, for example to influence Ukrainians about military service. Such material quietly appears on social media and gains thousands of views, making the problem particularly complex.
Microsoft has proposed a plan to authenticate online content. Its researchers evaluated various methods of documenting digital manipulation, including watermarks, digital signatures and file provenance tracing, and proposed standards that social platforms and artificial intelligence companies can apply. These tools don’t determine the truth, but show whether content has been altered.
Experts say that implementing such methods will make it much harder to spread deception, although it doesn’t completely solve the problem. People are often exposed to even known false content, but most want to know the truth. Some platforms have already started labelling artificially created content, but only partially, and full implementation of standards depends on companies’ business models and regulatory requirements.
Legislative initiatives such as California’s AI transparency law and EU regulations will oblige companies to disclose the origin of content. Microsoft is keen to participate in shaping such rules, while recognising that hasty or inconsistent labelling can undermine trust. While content verification tools are improving, it is important that they are applied carefully and consistently, given the complexity of today’s digital misinformation.