Watermarking and Traceability for High-Risk AI Outputs (Hong Kong)
Tags: Regulatory, Frameworks
TL;DR
- The Guideline calls for irremovable watermarks or embedded codes on high-risk AI outputs to ensure traceability and accountability.
- Examples include deepfakes, ID document images, and financial materials; developers/providers bear responsibility for accuracy and transparency.
Why it matters for HK marketers: Any synthetic personas, KYC assets, or financial communications must be provably AI-generated and verifiable to avoid legal and brand fallout.
Requirements and safeguards
- Traceability: Apply irremovable watermarks/embedded codes to high-risk outputs.
- Transparency: Train on verified sources; integrate trusted references; disclose information sources to users.
- Verification: Alert users when outputs cannot be verified; prevent spread of unverified content.
- Accountability: Independent evaluations and audits should apply to Technology Developers and Service Providers from the development stage.
3 high-risk output examples cited: deepfakes, ID documents, financial materials.
So what for marketers
Mandate watermarking/traceability features in your AI tools and verify them before campaign use—especially for identity, finance, or reputationally sensitive content.
← Back to Knowledge Base