Why aren't we recording pre-AI content using a PoE (Proof of Existence) protocol? (lobste.rs)

The post raises a simple but urgent question: as AI-generated media becomes indistinguishable from human-created content, why aren’t we exhaustively recording pre‑AI "proofs of existence" (PoE) — cryptographic timestamps or hashes — to preserve a verifiable baseline? This matters because PoE can provide a tamper‑evident record that a file existed at a certain time, helping researchers, journalists, and courts sort genuine pre‑AI artifacts from later forgeries and track provenance as synthetic media proliferates. Technically you can implement PoE in a few ways: compute a content hash (or HMAC for authenticated uploads) and anchor it in a centralized timestamp service or a decentralized ledger (blockchain) along with the transaction time. Practical toolchains already exist — RFC3161 timestamping, OpenTimestamps, content‑addressed storage (IPFS) with chain anchoring — but adoption is low because of friction: UX and cost, privacy/legal concerns about uploading raw content, trust and governance of centralized timestamp authorities, blockchain fees and bloat, key management, and the fact hashes don’t bind content to a human creator (you need digital signatures and device attestation for stronger provenance). For the AI/ML community the takeaway is clear: PoE can help, but it needs standardized workflows, signer‑bound keys, easy UX, and incentives to be effective before the pre‑AI window closes.