Overview

• The Wikimedia Foundation launched an opt-in, two-week trial on June 2 that placed AI-generated summaries from Cohere’s Aya model at the top of mobile Wikipedia articles to improve accessibility.
• Volunteer editors raised alarms over unverified content, fearing the summaries could introduce bias, undermine neutrality and erode trust in Wikipedia’s reliability.
• Facing a wave of negative feedback, the foundation paused the experiment on June 11 and announced a reassessment of its approach to AI integration.
• A project manager admitted the foundation had insufficiently engaged the Wikipedia community and pledged to enhance consultation before future trials.
• The foundation’s broader AI strategy, unveiled in April, reaffirms that no automated feature will be deployed without a robust, editor-led moderation workflow.