Computer

Protecting AI-driven technology still a challenge, say science and tech institutes

As artificial intelligence (AI) finds application in every walk of life, technology institutes in the country say that protection of AI-driven technology is a challenge.

In simple terms, AI is a field of computer science that trains machines to simulate the workings of the human mind. “Modern AI typically combines a deep learning architecture with a training algorithm, which is trained on task-specific data. This training yields a model for the task, which is then able to analyse any new data point and output its predictions,” said Mausam, professor and founding head of the School of Artificial Intelligence, at Indian Institute of Technology, Delhi (IIT-D).

Today, AI is the driving force behind technology interventions in the various fields such as education, healthcare, security and communication. While most technology institutes have empanelled lawyers for intellectual property (IP) rights, many are now seeking help from agents and consultants to handle the application process for AI.

The Foundation of Innovation and Technology Transfer (FITT) at IIT Delhi has now set up the Innovation Technology Transfer Office, which will help innovators and institutes in the field of IP. “There are many facilitators who can help in filing patents, but the quality of drafting the claims and strategising the IP protection is important which requires suitable expertise / guidance. We will be happy to help institutes with needful assessment and IP strategy through our new unit,” said Anil Wali, managing director, FITT.

Collectively, IITs file the largest number of patent applications, followed by institutes under the Council of Scientific and Industrial Research (CSIR). In a response to a question in Rajya Sabha, science and technology minister Harsh Vardhan had said that between 2009 and 2019, IITs had filed 3,751 patent applications and CSIR had filed 2,413 applications.

“About 50{c23a5fe915ca888c9672b1225e8f3bcb9450b05d8c59ef152c58a2c3a52fe784} of patents filed in the engineering field are from the software sector of engineering; which happens to be the most contentious and most litigious space in patent technology,” said Chirag Tanna, a registered patent agent and director of the firm Ink Idee, which offers consultancy services in patents, trademarks, copyrights, and design registrations. Among the institutes Ink Idee has worked with are IIT Bombay (IITB) and College of Engineering, Pune.

At IIT Delhi, around 3{c23a5fe915ca888c9672b1225e8f3bcb9450b05d8c59ef152c58a2c3a52fe784} of all IP applications are in the AI field. “This does not reflect the inherent strength which is substantial. Moreover, the Institute has also initiated a centre of excellence for AI and we expect increased inventive output in future,” said Wali.

“At present, roughly 2-3{c23a5fe915ca888c9672b1225e8f3bcb9450b05d8c59ef152c58a2c3a52fe784} of all patent applications are for technologies for AI-driven innovations. However, with AI now touching all aspects of life, we are seeing an increase in innovation in this field. It is a relatively new field and protection is a challenge. When it comes to AI, innovators generally find it easier to file for intellectual property (IP) rights in the United States of America or European countries,” said Milind Atrey, dean, research and development, IITB.

The European Patent Office (EPO) and US Patent and Trademark Office (USPTO) have a set of guidelines on how to examine patents relating to AI. “According to EPO, software and abstract mathematical models are not patentable, but if the idea is implemented, it then becomes eligible for patenting. In case of USPTO, one might be able to patent a specific implementation of an AI as long as there is a tangible effect,” said Tanna.

In India, algorithms can’t be patented under the Patents Act, 1970. “However, specific applications of AI-driven technologies for specific use-cases can very well be patented,” said Mausam.

“All software is abstract; so is AI – it is intangible, but its effect can be felt. Thus, it becomes very difficult to define the internal workings of IP in context of AI-driven technology. This is why many such patents get shot down early on. Secondly, computer-related inventions or software inventions have always struggled to sit within the stringent and finite definitions of law defined by Patent Acts across the world. The third is the challenge of authorship,” Tanna said.

AI-driven machines have the ability to learn from data and hold the potential to become autonomous, without needing the intervention of humans. “Existing intellectual property laws do not recognise an artificially intelligent machine’s right to invent a new piece of technology that can be patented; even though the world has reached a stage where artificially intelligent machines have, in fact, come up with new pieces of patentable technology,” said Tanna.

“Policy intervention needs to go hand in hand with innovation and challenges will always persist. However, rigorous testing of AI-driven technology can be a way forward. Each time the machine learns and evolves, it should be put to a careful and independent test,” said Mausam.

“Technical advancements always precede policies and laws. With an equitable dialogue with stakeholders, all issues can be resolved. These policies can encompass a variety of facets such as regulations on use of AI, ethics of use of AI in terms of decision making and decision enactments, burden of law when an AI makes decisions for and, on behalf of, human beings, boundaries of patentable subject matter when it comes to AI inventions and ownership and crediting of inventions made by AI inventions,” said Tanna.

app