AI’s invisible signature: Can the public verify what’s real?

Tech companies are adding hidden watermarks to AI-generated text and images. The goal is to help people tell machine-made content from human-made content. But researchers warn that the public cannot yet check these marks independently. Without independent scrutiny, watermarking may fail as a trust tool. Watermarking works by embedding a pattern in AI output. A detector then reads the pattern to confirm the source. Companies control both the pattern and the detector. Outside experts say this creates a verification gap: the public must trust the company’s own test. Independent public scrutiny means letting researchers, journalists, and civil groups test watermark systems. It requires open methods and shared standards. Supporters say this is needed to catch false positives and evasion. They also say it helps prevent watermarking from becoming a label that cannot be checked. The debate is timely. AI text and images are spreading fast in news, education, and politics. A mark that cannot be verified by outsiders may give false confidence. Experts say watermarking should support, not replace, media literacy and open reporting. The next step is clear: make verification public, or the mark means little.