---
title: "Cold Thoughts After the AIGC Carnival: Do We Still Need Copyright?"
author: "Unique Research"
sourcePublication: "Unique Research (FFCAP English library)"
originalPublishedAt: "2023-05-18T19:04:44+08:00"
canonical: "https://ffcap.cn/en/research/20230518a089pz00"
source: "https://ffcap.cn/en/research/20230518a089pz00"
language: "en"
cnSourceUrl: "https://view.inews.qq.com/a/20230518A089PZ00"
---

# Cold Thoughts After the AIGC Carnival: Do We Still Need Copyright?

_Unique Research · Insight / industry essay · May 18, 2023 · Full rewrite_

Author **Hanna**. Tech has no right/wrong — only trade-offs. Amid GPT-4 / ERNIE / Tongyi fever, an anti-AIGC wave gathers: human illustrators accuse AI of “theft”; some hope to poison training data. Unique Research spoke with GroupM China, Chopstick Tech, Unique Capital, and game veterans on commercialization beyond the noise.

## Does AI hold copyright?

Law anchors copyright in natural persons — AI out of scope. US scientist Stephen Thaler’s AI-image copyright bids (from 2019) repeatedly denied; pre-2023 rulings: AI lacks copyright. Chopstick CEO **Chen Wanfeng**: “Not so much no copyright as unprotected — not human-authored, no attribution.” Commercial AIGC may avoid classic infringement suits — but training/use can still touch protected works.

GroupM China brand-safety director **Jacob Zhou**: copyright should vest in the human user of the tool — copyright protects original expression of ideas, not ideas; training may hit protected expression; Prompt already injects human creative intent. Legislation pending.

## Infringement & liability

Even with open models + public-domain claims, creators remain angry; once trained, data is hard to unwind; DL opacity means developers may not know “how” the model thinks. Chen: training needs only high-res digital versions — like photographing Mona Lisa — hard to classify. If infringement assumed, liability may track human publisher (cf. AV rules). GroupM Choreograph **Vinny Ge**: AI is tool without rights/obligations; risk spans train→publish; need clear law + innovative licensing.

## Provenance

Web3 / smart-contract marking of authorized training (Chen) vs Unique Capital partner **Wu Wei**: “Tracing is hard — AIGC is a child recombining knowledge.” Brands should keep staged records to protect brand value. Deepfake/voice swap risk → Bilibili-style AIGC labels; China deep-synthesis rules require non-intrusive marks (Jacob Zhou) — else a “dark web” of unknown authorship.

## Trust & practice

Today AI content still minority online; inflection debated. Game vet “Cat-ge”: copyright fights are this generation’s ideology — later “must do, nobody cares.” Chen: 89% US college students use ChatGPT for homework; someday 99% of web may be AI — transition still needs labeling, and labelers will be AI (judge + athlete). Not AI untrusted — humans behind AI. Law must lead.

Embrace tools for design/campaigns (Fliggy, Meituan Mart, Zhong Xuegao AI ice cream). Don’t invent “AI projects” for PR — fold into brand concepts and packaging prototypes (Chen). Tools serve commercial outcomes. Expression still matters; intensity/frequency change with tools.

_Matrix red/blue pill close: face AI problems or ignore — tech has no morality; creators do._

— END —

---

Editorial English edition published on UniqueCapital: https://ffcap.cn/en/research/20230518a089pz00
On-site reading page: https://ffcap.cn/en/research/20230518a089pz00