[ REAL ML ]
[ In Person ]
AI Accountability, Decoded: Reflections on REAL ML 2026
Date
Location
Abstract/Summary
2026 workshop Co-Chairs Chenai Chair and Koliwe Majama reflect on bringing REAL ML to South Africa
In the midst of a brisk Johannesburg winter in June, we hosted the first African edition of the flagship REAL ML workshop, bringing together a diverse, interdisciplinary group of 35 activists, researchers, and creatives—largely representing the Global South—with the shared, resolute goal of establishing AI accountability frameworks that truly prioritize the needs of communities most affected by these evolving technologies.
The REAL ML workshop is a retreat, not a typical professional event. It creates space to drop competition, share unfinished work and blockers, and get peer support. This slower pace pushes back against today’s reactive AI culture, which often forces us to focus on a few powerful individuals, companies or trends over the experiences of significantly impacted communities.
Rooted in experience, committed to the field
Co-chairing the workshop was more than a contribution—it was a recommitment to the practitioner community, through facilitation, peer learning, mentorship, advice, and support.
Our work has long been shaped by capacity-building initiatives such as the African School on Internet Governance from the Association for Progressive Communications, and the Mozilla Foundation’s fellowship programs. Even amid uncertain funding cycles, sustaining institutional support for this work remains a priority.
Having both worked extensively in the digital rights field, we’ve seen how such programs build the critical mass our ecosystems need—fueling community-led innovation, strengthening leadership across major technology environments, and helping to ground technocrats in government in the lived realities resulting from impactful policy and technical decisions.
In a moment when civic freedoms and individual and collective rights are under increasing pressure, this kind of empathetic, accountable leadership isn’t optional—it’s essential.
That urgency was underscored by the turbulent digital rights landscape in Southern Africa, including the cancellation of RightsCon in Zambia the month before the REAL ML workshop, growing regional tensions, and stark visa disparities that mirror discriminatory border regimes worldwide.
Against this backdrop, the workshop held space for activists to surface shared vulnerabilities and collectively think through an evolving AI and technology landscape.
The overarching theme: Decolonizing AI and technology
A review of the 90 applications for this year’s workshop revealed a global pattern of algorithmic injustice—shaped by censorship, surveillance, and discrimination. This crisis of digital self-determination surfaced most clearly across three connected areas:
- Corporate power and its unchecked influence
- Language and cultural integrity in automated systems
- Human rights protections during political conflict and upheaval
The consensus is clear: today’s machine-learning ecosystems still prioritize scale over human safety.
For those of us living and working in the Global South, the current AI boom is also a moment to reflect on—and resist—a new wave of digital neocolonialism. Exploitation across global tech labor supply chains parallels the “algorithmic erasure” threatening Indigenous and minority languages. From the mines that extract raw materials, to manufacturing plants, to the moderation of sensitive content, the human cost is real. Accountability—from tech companies and governments alike—must value people and families over “smarter” machines or bigger bonuses.
Exclusion is never ‘accidental’
Through our week with the REAL ML community, we also saw from participants’ work and reflections at the workshop how digital exclusion consistently aligns with systemic discrimination. The harm caused by AI systems becomes normalized through everyday governance structures, cloud infrastructure ecosystems, and institutional decision-making. Ethical principles, human rights frameworks, and accountability mechanisms gradually erode as AI systems move from design phase to deployment, particularly in high-risk contexts such as conflict, elections, and public-sector use.
Biometric ID systems that marginalize certain identities and contribute to the loss of linguistic diversity are creating new forms of inequality. Turning these systems into tools for justice requires strong, locally relevant governance. Throughout the week, it became increasingly clear to us that building a more equitable future means shifting toward community-based, care-centered practices as an essential, not an optional, part of our work—especially as we try to mitigate localized harms such as intrusive data extraction and state surveillance.
Building a future forged in trust
The three-day agenda was meticulously co-designed to navigate through diverse formats, moving fluidly from open plenaries and solitary reflection to intensive thematic deep-dives, self-organized breakouts, a methodology crash course, and dynamic lightning talks. Beyond merely troubleshooting individual projects, the experience enabled us to construct a foundational scaffolding of mutual trust and profound respect for each person in the room.
By the final day, as each participant used five minutes to show how their work had shifted during the week, one thing was clear: even if the next points of collaboration were still uncertain, the gathering had done something rare.
With intellectual humility and a real commitment to radical inclusivity, the REAL ML workshop became a brief but vivid glimpse of the future we’re trying to build.
The impact on us
As leaders in this space who have moved through the iterations of technological development, being part of REAL ML was an opportunity to revitalise our belief that work can only be done in collaboration and solidarity with others. It impressed upon us what it means to be in a room where everyone is willing to learn from each other regardless of experience – hierarchies to the side. It also indicated the value of convening: finding fellow minded people helps to create vital spaces of rest and reflection in activism.
We are grateful to be part of the REAL ML community and we champion the need for spaces that allow people to learn, rest and glean hope for a future with technology that serves us.