I. Conceptual Framework: IP Protection in AI Governance

1. Forms of IP Relevant to AI Systems

AI systems used in courts (e.g., predictive sentencing, legal analytics tools) are protected through:

(a) Copyright

• Protects code, datasets (in some cases), and outputs
• Issue: AI-generated outputs may lack human authorship
• Courts often deny protection if no human creativity is involved 

(b) Patents

• Protect technical inventions and AI processes
• Requirements: novelty, non-obviousness, utility
• Problem: AI often seen as abstract algorithms, making patentability difficult 

(c) Trade Secrets (Most Common in AI)

• Protect:
  • Algorithms
  • Training data
  • Model architecture
• Advantage: No registration needed, lasts indefinitely
• Problem: reduces transparency in judicial systems 

2. Key Governance Tension in Judicial AI

AI in courts (e.g., bail prediction tools, legal analytics platforms) raises:

• Transparency vs Secrecy
• Fair trial rights vs proprietary protection
• Accountability vs innovation incentives

Trade secret protection often limits disclosure of algorithmic logic, which directly impacts due process rights

II. Detailed Case Laws (More than 5)

1. State v. Loomis (2016, Wisconsin Supreme Court)

Facts:

• Defendant challenged the use of COMPAS AI tool for sentencing.
• COMPAS algorithm was proprietary (trade secret).

Issue:

Can courts rely on a secret algorithm without violating due process?

Judgment:

• Court allowed use but imposed cautions:
  • AI should not be the sole basis of sentencing
  • Defendants must be informed of limitations

Significance:

• Recognized trade secret protection of AI
• But emphasized procedural fairness over IP secrecy

Relevance to Judicial Automation:

• Shows conflict between:
  • Corporate IP rights
  • Constitutional rights (fair trial)

2. R (Bridges) v. South Wales Police (2020, UK Court of Appeal)

Facts:

• Facial recognition AI used by police.
• Algorithmic details were not fully disclosed.

Issue:

Does lack of transparency violate fundamental rights?

Judgment:

• Use of AI was unlawful due to:
  • Lack of clear governance
  • Insufficient safeguards

Significance:

• Courts prioritized accountability over proprietary secrecy
• Implicitly limits IP protection where public rights are affected

3. Thaler v. Comptroller-General of Patents (DABUS Case, UK Supreme Court, 2023)

Facts:

• AI system “DABUS” listed as inventor in patent applications.

Issue:

Can AI be an inventor under patent law?

Judgment:

• Court held:
  ❌ AI cannot be an inventor
  ✔ Only natural persons can hold patents

Significance:

• Reinforces human-centric IP framework
• Raises governance issue:
  • Who owns AI-generated judicial tools?

Broader Impact:

• Affects:
  • AI-generated legal analytics tools
  • Ownership of automated judicial reasoning systems

4. Authors Guild v. Google (2015, US)

Facts:

• Google digitized books for search (Google Books).
• Used algorithms to analyze copyrighted works.

Issue:

Is algorithmic use of copyrighted data infringement?

Judgment:

• Held as fair use

Significance:

• Established:
  • AI systems can use copyrighted data for training
• Important for:
  • Legal analytics AI trained on case law databases

5. Waymo LLC v. Uber Technologies Inc. (2017, US)

Facts:

• Former Google employee allegedly stole self-driving car AI files.
• Uber used similar technology.

Issue:

Trade secret misappropriation in AI systems.

Outcome:

• Settled:
  • Uber paid ~$245 million equity

Significance:

• Reinforces:
  • AI algorithms = valuable trade secrets
• Shows risk in:
  • Judicial AI vendors handling sensitive data

6. Feist Publications v. Rural Telephone Service (1991, US)

Facts:

• Telephone directory listings copied.

Issue:

Are facts/data protected by copyright?

Judgment:

• No protection for raw data, only original arrangement.

Relevance to AI:

• Training datasets in legal analytics:
  • Raw legal data (case laws) = not protected
  • Structured databases = protected

7. Eastern Book Company v. D.B. Modak (2008, India, Supreme Court)

Facts:

• Copyright over Supreme Court judgments with editorial inputs

Issue:

Are legal judgments copyrightable?

Judgment:

• Raw judgments → ❌ not protected
• Edited versions → ✔ protected

Relevance:

• Legal analytics platforms rely on:
  • Case law databases
• Confirms:
  • Value lies in curation, not raw judicial data

8. American Geophysical Union v. Texaco Inc. (1994, US)

Facts:

• Corporate copying of scientific articles.

Issue:

Extent of fair use in commercial context.

Judgment:

• Not fair use.

Relevance:

• AI companies using legal data:
  • Commercial exploitation may limit fair use defense

III. Key Legal Issues Emerging from These Cases

1. Trade Secret vs Due Process

• State v. Loomis shows:
  • Courts allow secrecy but with limits
• Problem:
  • Defendants cannot challenge opaque algorithms

2. Ownership of AI Outputs

• DABUS case:
  • AI cannot own IP
• Implication:
  • Developers or institutions own judicial AI tools

3. Data Usage and Copyright

• Authors Guild v. Google:
  • Supports AI training on legal texts
• But:
  • Commercial misuse still risky (Texaco case)

4. Public Domain vs Private Control

• Eastern Book Company case:
  • Judicial data = public
  • Value-added analytics = private IP

5. Governance Gap

• No unified law for:
  • AI transparency
  • Algorithmic accountability
  • IP balance in judiciary

India especially lacks AI-specific IP clarity

IV. Application to Judicial Automation & Legal Analytics

Examples:

• AI bail prediction tools
• Case outcome prediction software
• Legal research engines

IP Challenges:

1. Algorithms hidden as trade secrets
2. Bias cannot be audited
3. Data ownership disputes
4. Unclear liability for wrong decisions

V. Conclusion

Protection of IP in AI governance for judicial systems is not just a legal issue but a constitutional one.

• Too much IP protection → opaque justice
• Too little protection → discourages innovation

Emerging Principle:

👉 “In judicial AI, transparency must override absolute proprietary control.”