Human authorship, strong copyright controls and potential funding losses were among the topics covered during the Standing Committee on Canadian Heritage’s final week of discussions on AI’s effects on the creative industries ahead of the body’s final report.

Warren P. Sonoda, national president of the Directors Guild of Canada (DGC), alongside director of policy Samuel Bischoff, testified that the organization’s members are concerned about the lack of guardrails around synthetic media.

He urged the government to reaffirm that the artist of a copyrightable work must be human, and added that, for a work featuring AI-generated elements, the human creator must have employed creative selection or composition or combined the AI elements with other elements to be eligible for copyright.

Sonoda also reiterated legislation and guidelines requests that were highlighted in earlier discussions, namely no text and data mining exceptions and mandatory transparency for AI-generated content.

Digital artist Eric Chan, professionally known as eepmon, argued more in favour of AI. Chan testified that the struggle between technology and rightsholders is not new, as artists have used technology — such as the printing press, the camera or the computer — to innovate for centuries, despite concerns that those emerging technologies would represent the end of art.

“The only thing that AI changes is scale, visibility and democratization of access,” said Chan. “Suddenly, the same remixing logic that powered culture for 300 years is labeled theft because a machine does it in seconds.”

While Chan acknowledged that many AI models use copyrighted material in their training, he said human artists do the same by studying and mimicking the music, novels and paintings of other artists in order to develop their own skills.

Chan argued that AI art should be protected through Canada’s fair dealing doctrine under the Copyright Act, which allows for the use of copyrighted materials without compensation as long as the output meets the criteria of research, private study, education, parody or satire.

“AI didn’t break copyright. It didn’t steal. It reveals our current system is functioning as it should,” said Chan. “Only now, accelerate it and democratize access so anyone can create and remix.”

Sonoda, while agreeing that artists build upon each other’s work, argued that AI is intrinsically different from that kind of innovation as it removes the human element from the equation.

Writers Guild of Canada (WGC) assistant executive director Neal McDougall testified that generative AI outputs are currently not copyrightable and that this should remain the case. He also argued that the level of human intervention required to make an AI work copyrightable must be high, otherwise “we risk the phenomenon of copyright laundering, in which financial incentives transform human beings into mere translators of machine outputs for the purposes of rendering AI outputs copyrightable.”

McDougall also raised concerns about the potential for funding from cultural agencies such as Telefilm Canada and the Canada Media Fund being allocated away from human creators and instead used to support AI-generated content.

Rezolution Pictures International president Archita Ghosh spoke to both the potential benefits and pitfalls of AI technology.

Ghosh said Rezolution uses AI daily for tasks such as structuring reports, research and administering file data. The company also used AI for its documentary Red Fever to generate an accurate image of a Haudenosaunee leader.

To accomplish this, Rezolution needed to train an AI platform with its own data in order to avoid biased data sets used by AI platforms that perpetuate stereotypes.

To combat biased data and increase Indigenous data sovereignty, Rezolution has begun a project, supported by SODEC, to create human-vetted and -curated data sets to improve Indigenous representation.

Despite this, Ghosh said that transparency on AI platforms is still lacking.

“The decades-long cultural and data research with which Rezolution trained the AI platform for Red Fever has disappeared into the AI abyss,” said Ghosh. “We hope it will be used for good, but whether it is or not, there is no footnote or reference connecting Rezolution and all the communities that contributed to it in our project with our curated data set. Our goal is to credit and compensate the researchers, content creators and cultural workers involved in creating AI-generated content.”

Pictured (L-R): Eric Chan, Warren P. Sonoda, Samuel Bischoff and Paul Fogolin

Image courtesy of ParlVU