> This includes increasing investment and research for technical solutions, such as machine detection and watermarking or cryptographic signatures, as well as wider adoption of other techniques, such as content provenance or hashing databases.

I don't see the point in including watermarks or signatures. A digital camera can include a ton of information that is completely removed and obfuscated by simply printing the photo and scanning it back in. The round-trip process of making anything analog and then digital again defeats almost any detection system.

PS, one of my two comments here was AI generated in its entirely by asking for a "Hacker News style" response. Did you immediately notice which one?

> This includes increasing investment and research for technical solutions, such as machine detection and watermarking or cryptographic signatures, as well as wider adoption of other techniques, such as content provenance or hashing databases.

If the same AI systems creating undetectable fakes are being used to detect them, we're essentially in an arms race where the house always wins. Watermarking only works if content creators widely adopt it, and bad actors obviously won't. Content provenance and hashing databases are interesting, but would require cooperation across platforms that have historically been reluctant to work together.

P.S. One of my two comments here was AI-generated by asking for a "Hacker News style" response. Did you notice?