The US Copyright Office’s latest decision to reject an application for an artificial intelligence-assisted artwork is a missed opportunity to offer more clarity on AI copyright issues.
The decision was possibly one of the first rulings involving an AI-assisted artwork since the office published its AI guidance in March.
Ankit Sahni, the applicant, had initially listed the AI tool 'RAGHAV Artificial Intelligence Painting App' as a co-author of the artwork.
But after receiving the first objection in February, he asked the US Copyright Office to list him as the sole author and to grant registration.
However, the office refused to reconsider Sahni’s request because the artwork failed to meet the level of creativity required to support a copyright claim in a derivative work.
I don’t want to get into the finer details of whether the office correctly applied US copyright law, but I do find some aspects of its decision disappointing.
Before getting into that, however, I must point out that the office is miles ahead of its global counterparts in terms of efforts to outline how AI-assisted works should fit within the existing legal framework.
At least two other intellectual property offices have granted Sahni’s applications and recognised AI authorship – in India and Canada (CIPO).
But India’s IP office tried to withdraw the registration, and the CIPO doesn’t conduct substantive examination and therefore did not assess the application on the merits.
The US Copyright Office, in contrast, has offered detailed rulings on at least three copyright applications for AI-assisted works.
In March, it published guidance aimed at harmonising examination practices.
Those efforts must be lauded.
Unfair review
You might argue, therefore, that it’s probably unfair to pick apart the decision to reject Sahni’s application after all the office's positive efforts.
But I hope my observations will help the authority develop a stronger AI jurisprudence.
My first problem with the decision is that the office’s decision to label the concerned artwork as a derivative at the appellate stage doesn’t make sense.
Of course, the office can give entirely new reasoning if previously undisclosed facts were brought to its attention during an appeal.
But the office’s first rejection in February, seen by Managing IP, suggested that it was already aware, or had at least gathered from the assertions made in the application, that part of the final output was derived from a photograph.
Despite that, the office still didn’t indicate at the time that this could render the final output a derivative work.
I’m not saying that classifying the artwork as a derivative work at the appellate stage was an afterthought, but the office does need to be more consistent.
Clarity needed
My second issue is that the office’s guidance gave hope to many stakeholders by indicating that some AI-assisted works could be registered if they met the level of human creativity required to award copyright protection.
What stakeholders would have liked from the Sahni decision was clarity on what exactly is the threshold of human creativity in the context of AI-assisted works.
In the absence of amendments to US copyright law, the office is, of course, allowed to determine the threshold and reject applications that don’t meet it.
But unless applicants have a better understanding of what meets the threshold, they are effectively taking a shot in the dark when filing or defending applications.
That’s not fair.
The lack of clarity also leaves scope for examiners to deliver discretionary decisions, contributing to imperfect jurisprudence.
That's not fair, either.
As the use of AI grows, IP offices will have to collaborate to decide how to accommodate it and other technologies in their legal frameworks.
For that to happen, IP offices must first bring some consistency and clarity to their own practices.