AI Code Ownership: Who's liable when your AI assistant copies code?

Not always. If your AI assistant just spat out those 200 lines of code, the law is still hazy on who the real "author" is. It might not be you, the AI, or even the company that built it. What a mess.

The issue of intellectual property is a legal minefield for code produced by artificial intelligence. Copyright laws, designed for humans, struggle to define a non-biological "author." In the United States, for instance, the U.S. Copyright Office has stated that works created solely by an AI are not protected by copyright.

This means if a model like GitHub Copilot or another AI assistant dishes out a piece of code for you, you might not have any exclusive rights to it. Anyone could use it without asking permission. Quite a gift, right?

This is where things get ugly. AI is trained on mountains of data, including lots of open-source or proprietary code. If it accidentally "learns" a protected snippet and reproduces it, you're the one in hot water.

The problem is that many AI models, like those used for code generation, have been trained on public repositories such as GitHub. These contain millions of lines of code with various licenses, some very permissive, others less so. It's not hard for a "learned" piece of code to be under an MIT license, but perhaps also GPL or other more restrictive ones.

If your AI generates code too similar to something under a GPL license, and you use it in a proprietary project, you could find yourself violating that license's terms. This exact risk led a group of programmers to sue GitHub Copilot in November 2022, citing copyright and open-source license violations. Who pays the bill? Usually, the company using that code.

Companies need to wake up and implement clear policies on using AI-generated code. Just hoping for the best won't cut it.

A careful review of AI-produced code is essential, perhaps using tools that check for originality and associated licenses. Some companies, like Google, have already started developing internal guidelines for their employees on the ethical and legal use of generative AI.

For the average user, the advice is simple: don't trust blindly. Treat AI-generated code as a suggestion, not gospel. Human oversight, for now, remains irreplaceable.