EIFL Public Library Innovation Award 20 for Public and Community Library AI Literacy Training Programmes

The Halcyon Africa Innovation in Agriculture and Food Security Accelerator supports early-stage, impact-driven ventures across Sub-Saharan Africa that are strengthening agriculture and food systems through market-based solutions. The 2027 cohort is open for applications until 23 October 2026. Technology is explicitly central to the scope: the programme welcomes tech-enabled agriculture and food security solutions across ten priority sectors, including agricultural data platforms, digital extension services, precision agriculture tools and agri-fintech innovations, all areas where AI and machine learning are typically the core of a competitive product. Each selected venture receives a 10,000 US dollar equity-free stipend and 10,000 dollars in Amazon Web Services credits, together with access to Halcyon's community of more than 650 founders, professional advisory support, expert-led workshops covering leadership development, capital strategy and product-market fit, and investor networking. The format is hybrid, combining virtual sessions with in-person residencies in Accra, Ghana and Nairobi, Kenya. What distinguishes this from most AI-for-agriculture funding is the applicant profile: it targets commercial ventures with working products and demonstrated traction rather than research teams, requiring paid-pilot or post-revenue stage with a working minimum viable product. For African AgTech founders building AI-driven advisory, monitoring or decision-support tools who have passed the prototype stage and need capital strategy support and investor access more than research funding, this is a well-matched and currently open opportunity.

The Bias Bounty Mapping Equity Challenge is a 10,000 US dollar competition run by Zindi with Humane Intelligence and support from the Heising-Simons Foundation, and it is methodologically distinctive in a way that matters to anyone working on AI fairness measurement. Rather than ranking submissions on aggregate accuracy, the challenge scores entries partly on how equitably a model performs across climate-vulnerable communities in the United States, operationalising the long-standing critique that a single accuracy number hides systematic underperformance on the populations most exposed to harm. The competition doubles as a pilot of Humane Intelligence's Bias Scoring API, so participants gain early hands-on exposure to a concrete fairness evaluation instrument rather than only a leaderboard placement, which is the main non-monetary reason to enter. The underlying task is geospatial mapping applied to climate and environmental equity, placing the challenge at the intersection of responsible AI evaluation and climate adaptation. Participation is open to all registered Zindi users with no geographic restriction on entrants, even though the dataset itself covers United States communities, which makes this one of the more accessible fairness-focused opportunities for researchers and practitioners outside well-funded institutions. Because it is a bounty-style competition rather than a grant, there is no proposal, no institutional affiliation requirement and no reporting burden; the submission deadline is 31 October 2026.

Foresight's AI for Science and Safety Nodes RFP is a small-cheque, high-variance programme aimed at work that conventional funders will not touch, and it is one of the few AI safety calls verifiably open with a future deadline of 31 October 2026. The framing joins two problems: ensuring a safe transition to highly capable AI, and using AI to unlock scientific breakthroughs. Proposals sit in one of three layers. Local compute covers community-owned and controlled computational infrastructure - an explicitly decentralist counterweight to frontier-lab concentration. Coordination and accountability covers human-AI interaction, supercollaboration and AI risk assessment mechanisms. AI-first science covers automated nanotechnology, whole-brain emulation and frontier biotechnology. The single most important screening criterion is stated plainly: Foresight wants projects where AI is the primary engine of progress, not projects where AI is, in its words, tapped onto an existing research program - so an established lab adding a machine learning component to ongoing work is the archetype of what gets rejected. Evaluation weights alignment with a focus area, impact on AI existential risk reduction, feasibility within short AGI timelines, execution capability, and high-risk high-reward potential. Two conditions are non-negotiable: all outputs must be open-sourced, and applicants must commit to active participation at Foresight's San Francisco or Berlin nodes, which makes this poorly suited to a purely remote team. Application is a single Airtable form with a roughly three-month review.