Introduction

The emergence of ChatGPT has sparked global interest in generative AI content (AIGC). From a legal perspective, AIGC raises complex questions across copyright, liability, data protection, and regulatory frameworks. This article examines the fundamental legal issues.

I. Copyright Issues

1. Does AI-Generated Content Qualify for Copyright Protection?

Chinese copyright law protects “original intellectual achievements of the human mind” expressed in certain forms. The key question is whether content generated by AI, without direct human creative input, can qualify for protection.

Current Chinese law does not explicitly address AI-generated works. The national copyright law revision draft has proposed acknowledging AI-generated content, but no final legislation has been enacted.

2. Ownership of AI-Generated Works

If AI-generated content qualifies for protection, who owns the copyright?

• The AI developer?
• The AI operator or platform?
• The user who provided the prompt?

3. Copyright Infringement in Training Data

AI systems are trained on vast datasets that may include copyrighted works. Questions arise:

• Does training on copyrighted works without authorization constitute infringement?
• Do copyright exceptions for computational analysis apply?
• What constitutes fair use in AI training contexts?

II. Legal Liability Issues

1. Liability for AI-Generated Content

When AI generates harmful, misleading, or illegal content, who bears responsibility?

• The AI developer for system design flaws?
• The operator for deployment decisions?
• The user for input choices?
• Or shared responsibility?

2. Product Liability

Could AI systems be considered products subject to product liability regulations if they generate harmful outputs?

3. Tort Liability

Operators of AI systems may face tort liability under general principles if AI-generated content causes harm to others.

III. Data Privacy and Security

1. Personal Data in Training

AI training datasets may include personal data. Compliance with China’s Personal Information Protection Law (PIPL) requires:

• Lawful basis for data processing
• Consent where required
• Data minimization principles
• Security obligations

2. Cross-Border Data Transfers

For AI systems with international components, cross-border data transfer requirements under PIPL and related regulations apply.

3. Data Security in AI Systems

AI systems must implement appropriate security measures to prevent unauthorized access, data breaches, and system manipulation.

IV. Regulatory Framework Development

1. Generative AI Regulations

China’s interim measures for generative AI services (2023) require:

• Compliance with laws and regulations
• Respect for intellectual property
• Accuracy and reliability of generated content
• Protection of personal information
• Security assessment and algorithm filing for certain services

2. Algorithmic Recommendation Regulations

AI content recommendation systems are subject to the “Regulations on the Management of Algorithmic Recommendations in Internet Information Services,” requiring transparency and user rights protection.

3. Future Regulatory Trends

Anticipated regulatory developments include:

• Specific AI legislation
• Enhanced transparency requirements
• Mandatory safety assessments
• International regulatory cooperation

V. Recommendations for AI Developers and Operators

1. Copyright compliance: Conduct due diligence on training data sources and licensing arrangements.

2. Content governance: Establish content review mechanisms and risk management systems.

3. Data compliance: Ensure PIPL compliance throughout the AI development and deployment lifecycle.

4. Transparency: Implement appropriate disclosure of AI-generated content and system capabilities.

5. Liability allocation: Clearly define contractual liability allocation with users, customers, and partners.