Tuesday, July 16, 2019

Protecting Artificially Intelligent Technologies

By Shantal Pai and Amanda McAllister

Artificial intelligence (AI) is becoming increasingly integrated into modern life. From self-driving cars to smart toothbrushes, AI can and will change the way we interact with the world around us.

The proliferation of AI has some inventors and designers considering how they can protect AI-related technologies through intellectual property rights. There also has been much discussion around how to build, design, and protect AI-related technologies without reinforcing biases and while still ensuring due process rights. 

As AI technology evolves, it has been less and less clear how intellectual property protections will apply to these technologies. For instance, most AI-related technologies are ineligible for patents. Courts refuse patents for “abstract ideas,” including software, and consequently, critical components of AI technologies, including data compilations and source code, may be ineligible for patent protection. 

Consequently, many have been looking to trade secrets as a potentially viable way of protecting rights in AI-related technologies. Trade secrets law protects economically valuable secrets when reasonable measures have been taken to protect their secrecy. In order to protect an AI-related technology as a trade secret, an AI developer must be able to do two things: (1) articulate what the trade secret is, and (2) take steps to protect the secrecy of the technology. 

Many developers are already adequately protecting the secrecy of their technologies. For example, appropriate measures to maintain the secrecy of an AI-related trade secret could include restricting access to or encrypting the code, using a VPN with a unique username and password, and requiring employees sign a proprietary information and invention assignment agreement. As AI continues to evolve with time, so too should the technologies designed to protect the AI. 

But articulating the precise trade secret can be a challenge for AI developers, as some AI technologies are capable of autonomously developing and evolving on their own. The developer may be unaware of the precise algorithm or process an AI uses to reach its conclusions. 

The ease with which other developers can circumvent protections, including from legitimate reverse engineering and independent development, poses a large risk to losing IP protections dependent on secrecy. If another developer is able to do either, the AI-related technology may no longer be eligible for trade secret protection. 

In addition to potential legal issues with intellectual property protections, the development of AI-related technologies also raises some ethical issues around due process and reinforcing biases. AI can integrate bias at each stage of the development process. These issues are compounded by the use of AI in decisions that affect individual rights and liberties, such as AI algorithms used in criminal justice decisions. Without full transparency in the decision-making process of AI, AI could actually amplify racial biases and undermine due process rights. Concerns about transparency of decision-making could affect multiple sectors, industries, and agencies, such as health care, government benefits, employment evaluations, insurance, and banking

Courts have just begun to grapple with these issues, and it is difficult to predict how case law might develop going forward. However, as AI becomes more and more integrated into everyday life, the law will need to balance intellectual property rights with the transparency required to preserve due process.

No comments :

Post a Comment