Recently, a phenomenon has been causing some concern—over 120,000 violations were taken down within 24 hours on a major platform, and the reason behind it is the misuse of large models to generate non-consensual fake images. This issue seems specific, but the underlying problems are actually quite common.

In simple terms, face-swapping technology can now be achieved with free tools, and the accuracy has reached a level that’s almost indistinguishable. This is nothing new, but the key point is—these tools have very low barriers to entry, and the dissemination cost is nearly zero. Even more troubling, there are already "customized generation services" targeting specific individuals on the dark web, with prices so low they’re absurd.

Victims want to defend their rights? It’s extremely difficult. The generated content has no authentic source, and while blockchain proof-of-ownership could theoretically solve this problem, the costs are prohibitively high for ordinary people. This creates a strange cycle: technological development outpaces policy-making, and platform measures like account bans only treat the symptoms, not the root cause.

There’s also a hidden risk worth noting—developers are beginning to illegally collect user privacy data under the guise of "AI training data needs."

This leads us to think: when the cost of creating fake content is far lower than the real-world harm it causes, should we reconsider our definitions of "evidence" and "truth"? Perhaps the idea of distributed proof-of-ownership on the blockchain really needs to be seriously considered, but the prerequisite is that the costs must be low enough for ordinary people to afford.