Thank you, Mr. Chair.
Dear Chair and members of the committee, good morning. Thank you for the opportunity to participate in this important work.
My name is Dave Forget, DGC's national executive director. With me today is Sam Bischoff, manager of policy and regulatory affairs. We appreciate the committee's invitation.
Generative AI threatens the ecosystem of creativity on an existential level. Creators should have the right to consent and be compensated whenever an AI entity uses their copyrighted content. As we stand at the crossroads of regulating AI to protect against harms, we believe it is crucial to protect creators from the economic and moral harms under AIDA.
The Directors Guild of Canada is a national labour organization that represents over 7,000 key creative and logistical personnel in film, television and digital media industries. Today it also includes over one thousand director members working across the country on screen-based programming.
The Canadian film and television sector generates massive amounts of value, employment and soft power. In 2021-22 the entire screen sector value chain directly contributed an estimated 337,000 jobs, $16.6 billion in labour income and $23.3 billion in GDP to the Canadian economy. However, artificial intelligence threatens the core of this ecosystem. Large language model developers are reproducing extensive amounts of creative works for commercial purposes without the authorization and fair compensation of authors.
Copyright remains a central framework law for governing our industry. Any unauthorized copying to train AI models is theft. Moreover, it is very difficult for rights holders to know when their works have been used without their consent in training AI models. Creators should be able to control whether their works are copied and used for mining purposes in the first place. Transparency in AI systems must be a prerequisite to defending authors' rights. This is a fundamental element to secure a future where human creativity can flourish.
In its current form, AIDA is failing to protect and uphold fundamental copyright principles. We need AIDA to do the following, consistent with the protections being provided to creators under the EU Artificial Intelligence Act, also known as EU AIA.
One, confirm that the use of copyright-protected content requires the authorization of the rights holder. This would be subject to the limited exception in copyright for technical ephemeral copying, which represents Canada’s very limited exception for text and data mining activities, to the extent that it applies.
Two, general-purpose AI systems like ChatGPT must be transparent about the materials used for training purposes. They should provide a description of information on the data used for training, testing and validation, as well as how this data was obtained and selected.
Three, providers of general-purpose AI models must be required to put in place a policy to respect Canadian law on copyright, including obtaining consent for text and data mining purposes. Any provider who makes available a general-purpose AI model in the Canadian market should comply with this obligation, regardless of the jurisdiction in which the AI training takes place.
I'll turn it over to Sam.