Sama Calls for Canada’s AIDA Regulation to Cover the Entire AI Supply Chain

We believe that Artificial Intelligence (AI) can be a force for good in the world with appropriate safeguards guiding the responsible and ethical development of AI. Like the regulatory inclusion of seatbelts and windshield wipers in cars, regulators need to embrace AI while ensuring individuals and communities are safe. We believe the current proposed legislation needs to be broader and cover the entire AI supply chain, from data acquisition to workers’ rights. The legislation should also include vetting and inspection processes for procurement by government entities.The Artificial Intelligence and Data Act (AIDA) is a key step in Canada’s leadership in the field of AI. We welcome this legislation and its companion document as an important step in regulating AI and ensuring that it is developed in a responsible manner and agree with other leaders in a call to pass this legislation. However, we will encourage regulators to continue to shed light on some of the details that remain unsettled. Regulation around Human Rights in the Supply Chain and incorporation of the OECD principles as a guideline for eventual regulation are key issues that have not yet been addressed.

Human Rights in the AI Supply Chain

An Ethical AI Supply Chain should include fair work standards for AI data annotation workers and acknowledge their value as a cornerstone of AI development. This is a missed opportunity for Canadian leadership in AI legislation.Why is this important? There is currently no AI legislation that protects workers in the AI supply chain. Canada has a history of being a leader in the inclusion of human rights principles in legislation and being at the forefront of ensuring that human rights are respected around the world.The failure to address this issue is leading to a race towards the bottom in the field of data annotation. Few platforms ensure that minimum wage laws are respected, as annotators are considered contractors. Many companies have business models that allow them to refuse work for any reason, leaving valuable annotators unpaid. Most annotation companies do not offer standard benefits packages or any form of guaranteed hours. Unless the playing field is leveled by applicable regulation or standards, the companies putting the most price pressure on the annotators will ultimately capture the market and push out those companies offering formal employment and protections to annotators.Therefore, the AIDA should include provisions to protect worker rights and ensure AI development is ethical and responsible. Regulation should require minimum standards on how data collectors and annotators are treated in the workplace, including regarding contract/employment, guaranteed minimum living wages, benefits and other considerations. Baseline standards should ensure all annotation suppliers consistently abide by the the principles in the EU Forced Labour Regulation. Additionally, and more urgently, government entities should be instructed to properly vet their data annotation suppliers to ensure they are complying with best labor practices.

Inclusion of OECD Principles

To codify international best practices and push for responsible AI, the Bill should align with OECD recommendations on AI. There are three key principles elaborated in the OECD Recommendation on Artificial Intelligence and further refined by the General Partnership on Artificial Intelligence and OECD. AI that are the cornerstone of how Sama operates, and should be the standard for all data annotation service providers:

• Human-centered values and Fairness: “AI actors should respect the rule of law, human rights and democratic values, throughout the AI system lifecycle. These include freedom, dignity and autonomy, privacy and data protection, non-discrimination and equality, diversity, fairness, social justice, and internationally recognised labour rights.”We lead the industry in ensuring every associate makes at the very least a living wage, including providing health benefits and all the protections that come with employment over contract work. Our employment model and end-point security means that training data isn’t sent to thousands of unverified and unidentified contractors.
• Inclusive growth, sustainable development and well-being: “Stakeholders should proactively engage in responsible stewardship of trustworthy AI in pursuit of beneficial outcomes for people and the planet, such as augmenting human capabilities and enhancing creativity, advancing inclusion of underrepresented populations, reducing economic, social, gender and other inequalities, and protecting natural environments, thus invigorating inclusive growth, sustainable development and well-being.”We are an active force for social justice through our social mission. We bring underrepresented voices to the table to ensure that data can be annotated by groups that otherwise wouldn’t have a seat at the table for spotting bias and blind spots in AI development. We give all workers a say in what work we take on through our Social Mission, and treat them fairly while we do it.
• Robustness, security and safety: “AI actors should ensure traceability, including in relation to datasets, processes and decisions made during the AI system lifecycle, to enable analysis of the AI system’s outcomes and responses to inquiry, appropriate to the context and consistent with the state of art.”We believe that robustness, security and safety for AI systems is far more achievable through an employment model for data annotators. Relying on disparate, unidentified contractors does not allow traceability or repeatability of results, and does not create feedback loops wherein the experience of annotators can be fully utilized to develop and refine annotation instructions, flag edge cases or alert developers to the flaws in AI systems.

We hope to make these practices standard in the data annotation industry as a whole.

Going forward, we will take a strong public position that reflects our social mission and values. Since our founding, we’ve pioneered new technologies and business methods that harness the power of markets for social good, meaningfully improving employment and income outcomes for those with the greatest barriers to formal work. Our model has employed over 13,000 people and helped over 65,000 people lift themselves out of poverty.

We want legislation that encourages all stakeholders in the AI supply chain, from data collectors to those deploying live AI systems, to build in a robust, secure and safe way that augments individual human and labor rights. Success in developing responsible and ethical AI systems requires accountability for the working conditions of data labeling associates. We welcome any vetting system that holds us and others accountable, as we have made it our mission to exceed these standards. Our work has already been recognized and validated via multiple awards on our impact, as well as accreditation from 3rd parties such as Poverty Action Labs, M.I.T., B-Corp certified, and Impact Real Leaders award.Our 15 years of experience, transforming from a non-profit to a Certified Benefit Corporation give us a unique position to showcase how AI can be used as a force for good.We are proud to be part of the Canadian AI ecosystem as it leads the way in AI legislation and regulation.Wendy Gonzalez, Chief Executive Officer, Alex Shee, General Manager of Canada, and Misha Benjamin, General CounselNote: We believe that Canada, the US, the UK and EU are leading the way in developing legislation that can have a positive and lasting impact on the development of Artificial Intelligence. In a series of blog posts, we will comment on each of these countries’ legislation individually, to state our position in context. Where possible, we will also formally send these to lawmakers in the hope that they include these positions in their legislation.