Digital Media Law

Key Concepts in this Chapter

What Is Artificial Intelligence?

• Broad term for technologies that perform tasks normally requiring human intelligence.

• Includes machine learning, natural language processing, computer vision, and generative AI (e.g., ChatGPT, DALL·E, Midjourney).

• Used in media for recommendation engines, automated journalism, video editing, and more.

Authorship & Copyright Issues

• U.S. Copyright Office: Works created solely by AI without human involvement are not copyrightable.

• Human authorship is required — but law is evolving around “human + AI collaboration.”

• Questions about ownership of training data and outputs.

Training Data & IP Rights

• Generative AI systems are trained on massive datasets, often scraped from the internet.

• Raises copyright, trademark, and right of publicity concerns.

• Litigation underway over whether AI training without permission infringes IP rights.

Bias, Transparency, & Ethics

• AI systems can reflect or amplify biases in their training data.

• Importance of explainability — understanding how AI reaches its outputs.

• Ethical debates over AI deepfakes, misinformation, and impersonation.

Liability Questions

• Who is responsible if AI-generated content infringes copyright, defames someone, or causes harm?

• Potential liability for developers, deployers, and users.

• Gaps in current law — few AI-specific statutes in the U.S. so far.

Regulatory Landscape

• U.S.: Mostly sector-specific or general laws applied to AI (IP, consumer protection, discrimination laws).

• EU: Moving toward comprehensive AI regulation (EU AI Act).

• Industry self-regulation: Voluntary commitments to transparency, safety, and responsible AI development.

AI in Media & Entertainment

• Automated scriptwriting, voice cloning, virtual actors.
• AI-assisted production tools for editing, animation, music composition.
• Concerns from creative professionals about displacement and credit.

Test Your Knowledge

1. Chapter 12 discusses that copying an author’s ‘style’ to guide an AI output:

 Constitutes per se copyright infringement

 Is prohibited by trademark law

 Is generally not protected by copyright as ‘style’ itself is not a protectable expression

 Is always fair use

2. According to Chapter 12, one practical newsroom safeguard against hallucinations is to:

 Rely on the model’s confidence score alone

 Publish immediately to preserve timeliness

 Disable source citations

 Require human editorial review (‘human‑in‑the‑loop’) before publication

3. Chapter 12 recommends or describes labeling/disclosure practices for AI‑assisted content primarily in order to:

 Comply with antitrust law

 Satisfy cable ‘must‑carry’ rules

 Promote transparency to audiences and editors

 Satisfy FCC sponsorship rules

4. The chapter notes that training may involve ingesting entire works. This consideration is most closely associated with which fair‑use factor?

 Factor 2 (nature of the work)

 Factor 1 (purpose and character)

 Factor 4 (market effect)

 Factor 3 (amount and substantiality)

5. On fair use, Chapter 12 frames the debate over training data as turning importantly on which factor when assessing market impact?

 No factor analysis is required

 Factor 1 (purpose and character) is the only relevant factor

 Factor 2 (nature of the work) only

 Factor 4 (effect on the market for or value of the original)

6. However, Chapter 12 notes that a human may claim protection for:

 Any prompt typed, regardless of creativity

 The human’s own selection, arrangement, or meaningful editing of AI outputs

 The AI model’s weights

 All raw training data on the internet

7. The chapter links AI ‘deepfakes’ and synthetic media primarily to legal risks involving:

 Patent non‑obviousness

 Broadcast sponsorship identification

 Right of publicity and potential defamation or privacy harms

 Cable retransmission consent

8. Chapter 12 characterizes ‘prompt engineering’ as:

 A process for registering copyrights

 A requirement under broadcast indecency rules

 A cybersecurity protocol in the SCA

 A method of refining instructions, context, and constraints to steer model outputs toward desired results

9. Bias discussed in Chapter 12 most directly arises when:

 A station fails to air E/I content

 A site lacks a privacy policy

 The model uses too much electricity

 Training data reflects skewed or unrepresentative patterns that the model then reproduces

10. An LLM, as discussed in Chapter 12, stands for:

 Large Language Model

 Limited Liability Mechanism

 Layered Latency Matrix

 Licensed Learning Module

11. According to Chapter 12, using a person’s distinctive voice in a synthetic ad without permission most directly raises:

 FCC licensing issues

 Only copyright concerns

 Only patent concerns

 Right of publicity/misappropriation concerns

12. In this chapter, ‘generative AI’ refers to systems that:

 Encrypt data for secure storage

 Transmit broadcasts over licensed spectrum

 Produce novel outputs (text, images, audio) by learning patterns from large datasets

 Merely sort existing files without creating new content

13. When completing a copyright application that includes AI‑generated portions, Chapter 12 emphasizes that applicants should:

 Register the underlying model instead

 Disclose and exclude the AI‑generated material while claiming only the human‑authored parts

 Claim the entire output without disclosure

 List the AI vendor as co‑author

14. The chapter uses ‘hallucination’ to mean that a model:

 Refuses to generate any output

 Encrypts all user data

 Refuses copyrighted inputs

 Outputs confident‑sounding but false or unsupported statements

15. A ‘prompt’ in the context of Chapter 12 is best described as:

 A command or instruction provided to the model to guide output

 A copyright registration code

 A broadcast station identifier

 A biometric template

16. Deepfakes that depict a real person making statements they never made are described in the chapter as posing which immediate newsroom risk?

 DMCA anti‑circumvention liability

 Defamation or false‑light exposure if published without verification

 FOIA exemption issues

 Patent invalidation

17. The chapter explains that the U.S. Copyright Office generally will NOT register a work:

 That is licensed under Creative Commons

 Created entirely by autonomous AI without human creative control

 That is published online

 Containing any quotations

18. Chapter 12 mentions the DMCA in connection with AI chiefly to note that platforms hosting user‑generated content can:

 Rely on §230 to avoid intellectual‑property claims

 Register the AI model with the Copyright Office

 Invoke §512(c) safe harbor if they expeditiously remove infringing material upon proper notice and lack actual knowledge

 Ignore all takedown requests

19. The chapter also notes that Section 230 of the Communications Decency Act:

 Applies only to broadcast stations

 Replaces the DMCA entirely

 Does not provide immunity for intellectual‑property claims and does not protect the platform’s own content

 Immunizes all AI outputs in every context

20. Regarding NDAs referenced in Chapter 12, pasting confidential client content into third‑party AI tools may:

 Be required for fair use

 Create a registered copyright automatically

 Strengthen confidentiality automatically

 Risk breaching NDAs or confidentiality obligations depending on the tool’s terms

 Loading …

Ideas for Future Study

• Human Authorship Standards: Should the law require a minimum threshold of human input for copyright?

• Licensing Training Data: Could a collective licensing system solve current disputes?

• Deepfake Regulation: Where should the line be drawn between parody and harmful deception?

• Labor Impacts: How should creative industries adapt to AI without undermining human talent?

• Global Coordination: How can laws keep pace with AI’s cross-border nature?

Parting Thought

AI is transforming the way we create, share, and experience media — but it’s also forcing us to rethink the very definitions of authorship, originality, and accountability. Should the law adapt to fit AI, or should AI be designed to fit within our existing legal frameworks?