Court of Milan: Machine Learning Does Not Infringe Patent, Declares Invalid
Case Law

Court of Milan: Machine Learning Does Not Infringe Patent, Declares Invalid

European Union·Briefly Analysis⏱️ 3 min read

Summary

  • The Court of Milan ruled on September 22, 2026, that a driver-assistance chip using machine learning did not infringe an obstacle detection patent.
  • The patent described a system comparing camera images with stored obstacle pictures.
  • The accused technology employed a machine learning chip to recognize obstacles.
  • The court found no infringement due to the distinct methods of operation.
  • Separately, the Court of Milan also declared the patent itself invalid.

What Happened

The court effectively distinguished between a rule-based, direct comparison system and an adaptive, learning-based system.

The Court of Milan's Specialised Business Section issued a significant ruling on September 22, 2026, in case RG 9864/2021 (decision No. 7233/2026). The court determined that a driver-assistance chip employing machine learning technology did not infringe a patent related to obstacle detection. Furthermore, the court independently declared the patent itself invalid.

This decision addressed a core question regarding the applicability of traditional patent claims to advanced artificial intelligence systems. The patent in question described a method and system for identifying obstacles in a vehicle's path by comparing real-time camera images against a database of stored obstacle pictures. The accused technology, however, utilized a driver-assistance chip that learned to recognize obstacles through machine learning, representing a fundamentally different approach to object identification.

The Technologies in Question

The patent at the heart of the dispute outlined a specific methodology for vehicle obstacle detection. This involved a system designed to capture images from a vehicle's camera and then perform a direct comparison of these live images with a pre-existing collection of stored images of known obstacles. The patent thus relied on a direct, image-to-image matching process to identify potential hazards.

In contrast, the technology accused of infringement operated on a distinct principle. It featured a driver-assistance chip equipped with machine learning capabilities. Instead of relying on a static database for direct comparison, this chip was engineered to learn and recognize obstacles dynamically. This adaptive learning approach differentiates it significantly from the more rigid, comparison-based method described in the patent.

Court's Determination

The Court of Milan concluded that the machine learning-based driver-assistance chip did not infringe the patent. This finding hinged on the fundamental difference between the patent's claim of comparing camera images with stored pictures of obstacles and the accused system's method of learning to recognize them. The court effectively distinguished between a rule-based, direct comparison system and an adaptive, learning-based system.

Beyond the non-infringement ruling, the Specialised Business Section also declared the patent invalid. This separate finding underscores that even if the technologies had been deemed similar enough to warrant an infringement analysis, the patent itself failed to meet the necessary legal requirements for validity. This dual outcome provides a clear resolution to the dispute.

Broader Implications

This ruling from the Court of Milan carries substantial weight for the patent landscape surrounding artificial intelligence and machine learning innovations, particularly within the automotive sector. It suggests a judicial inclination to carefully differentiate between patent claims that describe specific, rule-based processes and technologies that leverage adaptive, learning algorithms. The decision highlights the challenges of applying older patent frameworks to rapidly evolving AI methods.

The invalidation of the patent further emphasizes the need for robust and precisely defined claims when seeking protection for technologies that might interact with or be challenged by AI-driven solutions. For developers of machine learning systems, this outcome provides a degree of clarity, indicating that a patent based on a static comparison method may not readily encompass or block a dynamic, learning-based approach, especially if the patent itself is found to lack validity.

Source

Source: Original reporting via the Court of Milan

Get Deeper AI analysis

How does this affect you?

Get an AI analysis of this article grounded in your jurisdictions, practice areas, and any policy documents you've uploaded to Wansom.

Finish Reading the Full Story and the Expert Analysis.

Get the latest legal & regulatory intelligence in European Union

Instant access to full analysis, cited statutes & expert commentary
Customize your dashboard to track what matters to your business operations

Already have an account? Log in

Wansom is AI and can make mistakes.