Protection Of AI-Assisted Software Architecture In Irish Patent Practice.
✅ Protection of AI-Assisted Software Architecture in Irish Patent Practice
Ireland follows European Patent Convention (EPC) principles. Therefore, Irish patent practice is heavily shaped by EPO case law. Since there are no major Irish court decisions specifically on AI-assisted software architecture, the EPO case law is used as the primary guide.
🔎 Key Legal Principles (Irish Patent Practice)
1. Patentable Subject Matter
Under Irish law, a patent is only granted if the invention is:
New
Involves an inventive step
Industrially applicable
Not excluded from patentability
2. Main Exclusion for Software
Computer programs “as such” and mathematical methods are excluded.
Therefore, AI-assisted software architecture is not automatically patentable. The software must deliver a technical effect or solve a technical problem.
🧠 What is AI-Assisted Software Architecture?
AI-assisted software architecture means using AI to:
generate code
design system components
optimize system performance
auto-configure systems
propose architecture patterns
Examples:
AI that generates microservice architecture from requirements
AI that optimizes database indexing
AI that designs security policies automatically
⚖️ Core Question in Irish Practice
Is AI-assisted software architecture a “computer program as such”?
➡️ If yes → Not patentable
➡️ If it produces a technical effect → Potentially patentable
🧩 Case Laws (More than Five) — Detailed Explanation
1. T 1173/97 (IBM) – “Computer Program Product”
Facts
A claim covered a computer program product stored on a carrier.
Holding
A computer program is not excluded if it produces a further technical effect beyond normal computation.
AI Architecture Implication
AI-assisted architecture could be patentable if it produces a technical effect like:
reduced latency
improved resource utilization
automated system control
2. T 641/00 (COMVIK) – “Technical Features Only”
Facts
The invention included technical and non-technical features.
Holding
Only technical features contribute to inventive step.
AI Architecture Implication
AI-generated architecture alone is non-technical.
But the architecture may include technical features such as:
improved resource scheduling
optimized data flow
automated error recovery
Only those parts can support inventive step.
3. G 1/19 – “Simulation and Technical Character”
Facts
A simulation claim was questioned for technical character.
Holding
Simulation can be patentable if it provides a technical effect.
AI Architecture Implication
If AI architecture is used to simulate:
system performance
network loads
energy consumption
and produces technical improvement, it may be patentable.
4. T 1784/06 – “Medical Image Processing”
Facts
An image processing algorithm for medical images was claimed.
Holding
The algorithm was patentable because it produced a technical effect on real-world imaging.
AI Architecture Implication
AI-assisted architecture for:
real-time embedded systems
industrial control systems
safety-critical systems
may be patentable if it improves real-world performance.
5. T 1227/05 – “Traffic Management”
Facts
A computer-based traffic management system was claimed.
Holding
Patentable because it improved traffic control, a real-world technical effect.
AI Architecture Implication
AI-assisted architecture that improves:
cloud orchestration
traffic routing
data center management
could be patentable if it has a real-world technical effect.
6. T 2047/12 – “Technical Effect from Data Processing”
Facts
The case dealt with data processing and whether it produces a technical effect.
Holding
Data processing can be technical if it improves hardware or system performance.
AI Architecture Implication
AI-assisted architecture that optimizes:
CPU usage
memory consumption
network throughput
may be patentable.
7. T 1457/14 – “Neural Networks and Technical Effect”
Facts
A neural network was claimed for pattern recognition.
Holding
The neural network must solve a technical problem.
AI Architecture Implication
AI architecture that only proposes design patterns is not enough.
But architecture that improves system reliability or reduces failure may be patentable.
8. T 1952/16 – “Training Optimization”
Facts
A method claimed optimized AI model training.
Holding
Rejected because it only improved internal computation, not a technical effect.
AI Architecture Implication
AI architecture that improves training speed alone is not patentable unless:
it reduces power consumption in hardware
improves real-time system performance
⚙️ What Can Be Patented in AI-Assisted Software Architecture?
Here are patentable AI-assisted architecture scenarios in Ireland:
✅ 1. AI Generates Architecture for Real-Time Control Systems
Example:
AI designs architecture for robotics control.
Technical effect: improved reaction time and stability.
✅ 2. AI Optimizes Embedded Systems
Example:
AI automatically partitions tasks between CPU and FPGA.
Technical effect: reduces latency and power consumption.
✅ 3. AI-Assisted Security Architecture
Example:
AI designs security policies and firewall rules.
Technical effect: improved security performance and reduced breaches.
✅ 4. AI Architecture for Cloud Resource Management
Example:
AI auto-scales microservices.
Technical effect: reduces cost and improves uptime.
❌ What is NOT Patentable?
❌ AI that only generates design patterns
❌ AI that outputs architectural diagrams
❌ AI that only automates software development
❌ AI that only improves coding speed
These are considered business or administrative methods, not technical.

comments