Mila AI Policy Fellowship

Information integrity—the ability to access, verify, and trust knowledge—is a cornerstone of democratic decision-making, economic activity, and scientific progress. The rapid proliferation of large generative models and agentic AI systems is fundamentally transforming our information ecosystems. This is offering unprecedented opportunities for knowledge creation while introducing systemic risks of large-scale manipulation, automated misinformation, and the erosion of institutional trust. This thematic area focuses on developing actionable strategies to advance trustworthy knowledge ecosystems and safeguard the social contract in AI-mediated environments. Proposals will align with principles of democratic governance, such as ensuring the quality of civic interactions and opportunities for political participation alongside fair and integral democratic elections, establishing institutional accountability and upholding the rule of law, as well as promoting human rights and substantive equality. These policy challenges require a clear understanding of the capabilities and limitations embedded in the technical design of models, platforms, and digital infrastructures.

Example Topics:

• Trust and public discourse in the age of synthetic media: Developing policy standards for watermarking, provenance, and auditing in multimodal models to sustain editorial standards and public-interest media.
• Electoral integrity and agentic systems: Addressing the policy challenges posed by the use of generative AI in campaigns, including targeted persuasion, coordinated manipulation, and the rise of autonomous misinformation bots.
• Algorithmic amplification and information flows: Investigating regulatory approaches to recommender systems and their role in shaping public discourse, mitigating "engagement-at-all-costs" models.
• Institutional accountability and contextual liability: Charting pathways for effective legal and accountability frameworks that account for AI opacity, autonomous behavior, and the limited predictability of continuously adapting systems.
• Rights-centered AI policy: Revisiting legal frameworks to ensure the protection of freedom of expression, privacy, and the right to information (access to data) within AI-driven systems.
• Economic information and financial integrity: Addressing the reliability of corporate and market information in the face of AI-driven financial fraud, market manipulation, and automated economic decision-making.
• Democratic alternatives to centralized AI: Exploring "public-interest AI" models and constitutional AI frameworks that prioritize collective input and democratic alignment over centralized private ownership and control.
• Health and science communication: Delivering policy responses to large-scale AI-generated misinformation in health and science communication, including the credibility, accessibility, and governance of reliable public knowledge.

As AI systems become deeply integrated into our lives, acting at times as the primary delivery of healthcare services, digital social spaces, and professional environments, they present unique opportunities for innovation alongside significant risks to mental health and psychosocial wellbeing. This thematic area explores the profound interplay between AI systems and the physical, mental, and social dimensions of human health, with a primary focus on protecting the rights and wellbeing of youth and workers. Safeguarding the cognitive and social development of younger generations while ensuring that the modern workplace remains adapted to human needs, this area focuses on ensuring that technological adoption promotes holistic wellbeing, psychological safety, and meaningful human-centric growth.

Example Topics:

• Mental health and human-AI interaction: Designing interventions to prevent harm and promote psychological safety in environments where AI mediated communication or decision-making is prevalent.
• Youth development and children’s rights: Addressing the long-term cognitive and developmental impacts of human-AI interactions on younger generations and establishing protections for children’s digital rights.
• Biologically responsive and responsible AI (policy): Exploring regulatory standards or obligations that reflect intentional behavioral and cognitive impacts of human-AI interaction (engagement, dependency, cognitive “off-loading”).
• Modernizing labour health and safety standards: Revising existing policies to address psychosocial harms that arise alongside AI integration in work environments, such as shifts in job control, erosion of social support, and impacts on job satisfaction.
• Institutional participation and worker voice: Developing policy suggestions for enabling organized labour and worker representatives to be active participants in the design, adoption, and governance of workplace AI. Can equally apply to health and mental health care professionals employing AI for health care delivery. 

Amidst global geopolitical shifts, the centrality of AI sovereignty and security have emerged in national and international fora. This thematic area navigates the complexities of competing narratives and definitions of AI sovereignty, and real-world security implications inherent in the convergence of global innovation and critical infrastructure. As AI integration scales across public and private sectors, new risk vectors emerge - from AI supply-chain vulnerabilities, to interoperability issues, and large-scale cybercrime. This track focuses on shedding light and finding solutions to maintain integrity and stewardship over their data and resources while enhancing systemic resilience in an increasingly digital and interconnected world. 

Example Topics:

• Research security and supply chain integrity: Developing guidance for safe AI development that considers the origin of source code and the security of key supply chain inputs.
• Cyber-resilience and critical infrastructure: Ensuring AI contributes to the security of essential services rather than introducing new vulnerabilities, for example through procurement.
• Data commons and shared infrastructure: Exploring ways for nations to combine compute, talent, and data resources to challenge centralized ownership models.
• Energy and compute sovereignty: Addressing unequal and volatile access to the energy and computing power that underpins AI advances.
• Interoperability and mutual recognition: Identifying cross-jurisdictions assurance mechanisms that promote meaningful international collaboration in academia, R&D and trade.
• Data sovereignty and stewardship: Balancing national priorities and across multiple jurisdictions and global innovation supply chains, ensuring integrity of data ownership and stewardship.

Indigenous AI centers Indigenous sovereignty and self-determination at the intersection of artificial intelligence, focusing on how communities steward and advance their own cultures, languages, and knowledge systems. As AI increasingly reshapes global governance, economic development, and digital infrastructure, there is an urgent need to ensure that Indigenous Peoples—in the Canadian context and beyond—are AI policy leaders and decision makers. This thematic area prioritizes the development of policy frameworks that move beyond simple harm mitigation toward the realization of Indigenous data sovereignty and stewardship. 

Suggested Topics:

• Indigenous data sovereignty and governance: Developing policy mechanisms for the ownership, control, and stewardship of data related to Indigenous peoples, lands, and resources within national and international AI strategies.
• Collective privacy and consent in Machine Learning: Addressing limitations of individualistic privacy laws by designing frameworks for collective data privacy and community-based consent in the training of large-scale models.
• AI for language and cultural revitalization: Proposing guidelines and technical standards for the use of AI in preserving Indigenous languages while safeguarding against the unauthorized extraction or commercialization of traditional knowledge.
• Ethical AI adoption and impact mitigation: Developing community-led impact assessment tools to evaluate the socio-technical effects of AI deployment within Indigenous territories and governance structures.
• Infrastructure and economic inclusion: Policy pathways for strengthening Indigenous leadership in the digital economy through investments in community-owned digital infrastructure, entrepreneurship, and specialized talent development.
• Indigenous AI knowledge systems: Integrating Indigenous epistemologies into the design of AI ethics and safety protocols to foster a more diverse and representative global AI landscape.

As AI enables hyper-personalized learning and necessitates large-scale reskilling, policy guidance must evolve to manage systemic transitions while preventing harm and ensuring inclusive access. This thematic area addresses the transformative role of AI in education  and learning environments, as well as its profound implications for literacy, equity, and the future of the global workforce. The focus is on fostering beneficial change management and organizational transformation—from primary education to professional development—to support lifelong learning. This track prioritizes the protection of student and educator data rights while exploring how AI can be leveraged to enhance human-AI co-reasoning and epistemic skills, ensuring that no community is left behind in the transition to an AI-driven economy.

Suggested Topics:

• Equity and harm prevention in personalized learning: Designing governance frameworks to ensure AI-driven educational tools do not exacerbate existing socio-economic inequalities or introduce biased pedagogical outcomes.
• Privacy and data protection in education: Establishing rigorous standards for the collection, use, and long-term protection of student and educator data within AI-enabled learning platforms.
• Workforce reskilling and career transitions: Developing policy insights for interventions to address displacement and ensure labor market readiness through inclusive, large-scale reskilling and upskilling pathways.
• AI literacy: Proposing roadmaps for enhancing and fostering critical AI literacy skills within relevant tiers of education, higher education, and continued learning.
• Worker voice in educational technology: Ensuring that educators, organized labor, and student representatives are active participants in the design and adoption of AI systems within schools and workplaces.
• Human-AI co-reasoning and epistemic skills: Exploring policy supports for pedagogical shifts that emphasize critical thinking, media literacy, and cognitive resilience in an era of automated knowledge production.
• Inclusive institutional participation: Developing strategies for systemic transformation that empower marginalized communities to lead in the adoption of safe and responsible AI within educational institutions.

AI offers powerful tools for climate action, yet it bears its own environmental footprint. This thematic area focuses on policy solutions at the intersection of AI, climate, and the natural world. It encompasses policy initiatives aimed at the management of AI’s own environmental footprint, alongside the promotion of high-impact applications that leverage AI for environmental protection, climate adaptation, and disaster mitigation. Proposals may also focus recommendations for AI utilizations to support efficiency, innovation, and stewardship across natural resource sectors, including forestry, mining, and agriculture, fisheries, water management, and energy systems, and more. 

Example Topics:

• Environmental footprint and e-waste: Transparency standards for AI's energy, water, and material use, alongside management guidelines for sustainable hardware lifecycles (procurement, circular economy, disposal practices etc.).
• Fit-for-Purpose AI ("Small AI" vs. "Big AI"): Guidance on selecting appropriately scaled, resource-efficient models for specific industrial or environmental tasks.
• Natural resource management: Identifying strategic support for AI applications for productivity and stewardship in critical mining, forestry, agriculture, and fisheries, etc.
• Climate adaptation & disaster mitigation: Insights for AI deployment enabling early warning systems, emergency response planning, and infrastructure resilience.
• Energy systems: AI deployment for more efficient electrical grids, load balancing, and broader energy savings.
• Climate misinformation: Interventions to safeguard scientific consensus in AI-mediated information ecosystems against climate-related deception. 

We welcome well-developed, thoroughly considered,  interdisciplinary, and policy-relevant proposals aligned with one or more of the thematic areas and the objectives of the Fellowship. The proposal must:

• Clearly demonstrate what you seek to achieve during the Fellowship. This includes describing the project’s key objectives with specific policy issue(s) or gap(s), target audience, and intended policy impact;
• Align with the selected thematic area, or clearly demonstrate the relevance of other proposed topics; 
• Demonstrate how you seek to translate research into actionable policy insights.
• Introduce original, independent work; heavily AI-generated proposals will make it difficult to assess your fit;
• Demonstrate feasibility within the six-month fellowship period.

Maximum Length: 1,000 words / corresponding 2 pages (excluding references, if applicable).

Yes. Applicants may submit proposals outside the predefined areas, but should clearly explain the relevance to AI policy, feasability, and potential for impact. 

Yes. The Fellowship will pair you with a Mila advisor and a member of the policy secretariat. However, you may also propose other collaborators in your application, including funders and institutional partners.

Fellows will be selected based on a number of criteria. These include a thorough evaluation of the quality and potential for impact of submitted project proposals, alignment with thematic areas or relevance of topics other than listed, pairings with advisors and policy secretariat, as well as an overall consideration of diversity of cohort. Shortlisted candidates may be invited for a brief interview before final decisions are made.

The working languages of the fellowship are English and French.

No. The program is open to applicants world-wide. The Fellowship is delivered in a hybrid format, with in-person activities scheduled in Montreal, Quebec. Visa support can be provided for selected fellows who need it.

Yes. If you are selected and require a visa to attend the in-person component, Mila can support your application process.

Yes. Compensation is based on an hourly rate for a dedicated 15 hours per week. The rate is variable within a set range based on profile and experience.

No. There are no application fees. Participation is fully funded, including travel and event costs for  in-person components

Yes. All travel and event-related costs associated with the required in-person components of the fellowship are covered by the program.

We strongly encourage live participation in key sessions. We aim to accommodate fellows in different time zones and core sessions will be scheduled with input from the cohort to maximize inclusivity.