The 404 Media Podcast

Episode: "How Wikipedia Will Survive in the Age of AI (With Wikipedia’s CTO Selena Deckelmann)"

Date: January 19, 2026
Host: 404 Media Team | Guest: Selena Deckelmann, CTO of Wikimedia Foundation

Episode Overview

This episode celebrates Wikipedia's 25th birthday by exploring how one of the world's most important online encyclopedias plans to continue thriving amidst the rise of generative AI. Host(s) from 404 Media are joined by Selena Deckelmann, Chief Product & Technology Officer at the Wikimedia Foundation, to discuss Wikipedia’s unique model, community governance, resilience, and the challenges and opportunities AI brings to the platform.

Key Discussion Points & Insights

1. Defining Wikipedia: Website, Community, and Process

[00:00–04:20]

• Wikipedia as a Project: Selena emphasizes Wikipedia as more than a website—it's a massive collaborative effort with 250,000 editors, 65 million articles, and over 300 languages, serving 15 billion page views per month.
• "The Edit Button": The ability for anyone to edit ("a webpage with a button that just says edit") is central and surprisingly sustainable due to community-driven checks and balances.
• Quote:
  "When I think about what it is, it's that button, it's that edit button that anyone can press. It's all of these people who have worked together to produce the corpus of knowledge ..." (Selena Deckelmann, 03:12)

2. Community vs. Foundation: Roles and Relationship

[04:20–08:24]

• Editors & Foundation: Editors self-organize and create content, while the Wikimedia Foundation provides technical, legal, and infrastructure support.
• Distinct, Yet Overlapping: The lines between community (content creation) and foundation (support systems) are clear but collaborative, with tens of thousands of technical volunteers as well.
• Notable Example: "Project Tropical Cyclone" organizes real-time editorial coverage during disasters.

3. Communication & Governance: Open, Complex, and Human

[08:24–13:53]

• Thousands of Channels: Communication happens across IRC, Telegram, Discord, mailing lists, and more.
• Advisory Structures: The Product and Technology Advisory Council brings experienced volunteers together to advise on infrastructure and process improvements.
• Conflict Management: Open systems foster frank discussion but also conflict; research indicates that cross-perspective collaboration (e.g., editors of differing political backgrounds) leads to greater neutrality over time.
• Quote:
  "...over time when folks [with opposing political views] do that, their editing becomes more neutral... as a result of being exposed to these different ideas." (Selena Deckelmann, 12:23)

4. Human Judgment & Neutrality in Content

[13:53–15:17]

• Editorial Groups as Clearinghouses: Groups like the Cyclone editors play crucial roles in vetting information during emergencies.
• Reliability Amid Shifting News: Wikipedia's editor base routinely sifts through conflicting reports to provide accurate information.

5. Wikipedia and AI – Principles and Concerns

[15:17–23:24]

• Human-Led Ethos: Despite integrating some machine learning, Wikipedia remains fundamentally a human-led project.
• Sustainability & Reciprocity: Selena stresses the web’s original system of reciprocal links, now challenged by AI that doesn't inherently link back to sources.
• Potential Crisis: AI systems, particularly LLMs and chatbots, often aggregate info without supporting or referring users back to Wikipedia, threatening the ecosystem that sustains the encyclopedia.
• Training Data Dependency: Studies show LLM outputs become more toxic without Wikipedia in their training data, highlighting Wikipedia’s qualitative impact on the broader information ecosystem.
• Quote:
  "...when you remove Wikipedia [from data], the responses [from the language model] suddenly become much more toxic. And I think that's so interesting..." (Selena Deckelmann, 22:01)

6. Commercial Partnerships & Sustainable Infrastructure

[27:59–33:47]

• Partnerships with AI Companies: Wikipedia works with companies like OpenAI and Google via Wikimedia Enterprise APIs, seeking support for infrastructure rather than pure data exploitation.
• Scraping Problems: Unsanctioned bots, scraping at scale (sometimes via residential proxies), create technical and sustainability challenges, sometimes amounting to denial of service attacks.
• Need for Alignment: The foundation aims for partnership models that reinforce the commons and redistribute value back to content creators.

7. Machine Learning Inside Wikipedia

[33:47–37:54]

• Supporting Tools: Machine learning (ML) is used for edit analysis (e.g., revert risk scores), grammar/spelling checks, citation prompts, and "peacock language" detection.
• Empowering, Not Replacing Editors: The focus is on augmenting human editors’ productivity, not replacing them.
• Quote:
  "How can we make an individual editor that much more effective in a moment, make them more powerful?" (Selena Deckelmann, 36:02)

8. Detecting and Addressing AI-Generated Content

[37:54–42:47]

• WikiProject AI Cleanup: Volunteers spearhead efforts to find and remove erroneous AI-generated content, with the foundation supporting with tooling and research.
• Community-Led, Foundation-Supported: Foundation’s approach is collaborative and adaptive, providing detection tools and responding to community requests rather than unilateral intervention.

9. Governance, Attitudes Toward AI, and Internal Divergence

[42:47–49:52]

• Tensions Around AI: Some editors strongly resist AI-generated content or features (like AI summaries), while others (including Jimmy Wales, co-founder) are more open to machine learning as an assistive tool.
• Neutral but Practical: Deckelmann advocates for pragmatism—testing and deploying ML when it genuinely helps, but always in dialogue with the community.
• Quote:
  "I don't think that we should just ban this particular technology ... it's complicated ... and I think everybody sees that it's a complex situation." (Selena Deckelmann, 48:51)

10. The Future of Wikipedia & The Internet

[49:52–53:01]

• "Lightning Round" Futures:
  • 100 years from now? "Yes..." (Selena Deckelmann, 50:09)
  • Always a human in the loop? "Yes, but..." (Selena Deckelmann, 50:16)
• Broader Implications: Wikipedia's governance model is more relevant than ever, and its “magical” community-led structure could inspire more online projects.
• Necessity of Transparency, Humor, and Good Faith: The key ingredients to Wikipedia’s resilience are assumed good faith, humor, and a willingness to deal with the unexpected.
• Quote:
  "I think more of [the Internet] can be done ... that's the way that we will produce more systems like this. And I absolutely think it's possible ... even if it's hard." (Selena Deckelmann, 52:25)

Notable Quotes & Memorable Moments

• [03:12] "It's that button, it's that edit button that anyone can press ... that creates such an incredibly reliable, verifiable, transparent system for knowledge." – Selena Deckelmann
• [12:23] "...over time ... their editing becomes more neutral, likely as a result of being exposed to these different ideas." – Selena Deckelmann
• [19:35] "...an agentic AI system, I'm not really sure where the reciprocity is. That's the big question..." – Selena Deckelmann
• [22:01] "When you remove Wikipedia, [LLM] responses suddenly become much more toxic. And I think that's so interesting..." – Selena Deckelmann
• [36:02] "How can we make an individual editor that much more effective in a moment, make them more powerful?" – Selena Deckelmann
• [48:51] "I don't think that we should just ban this particular technology ... it's complicated ... and I think everybody sees that it's a complex situation." – Selena Deckelmann
• [52:25] "I absolutely think it's possible ... even if it's hard." – Selena Deckelmann

Important Timestamps

• 00:00–04:20 — What is Wikipedia? Mission, scale, and the meaning of the "edit" button.
• 08:24–13:53 — Communication channels, advisory councils, and collaborative neutrality.
• 15:17–23:24 — AI’s impact, Wikipedia’s role in LLM training data, and reciprocity crisis.
• 27:59–33:47 — Wikipedia's relationship with major AI companies, infrastructure strain, and sustainable partnership.
• 33:47–37:54 — ML inside Wikipedia: revert risk, peacock language, content assistance.
• 37:54–42:47 — Detecting/removing AI-generated content, ISBN fraud, and community process.
• 42:47–49:52 — Community vs. foundation tensions, philosophy around blanket AI bans, and practical neutrality.
• 49:52–53:01 — "Lightning round" on Wikipedia's long-term future and the potential for more Wikipedia-like platforms online.

Summary & Takeaways

• Wikipedia’s Resilience: Its open, human-moderated model, collaborative governance, and tools for accountability continue to make it unique and essential in the AI era.
• AI as Both Threat & Tool: Generative AI challenges Wikipedia’s ecosystem (especially reciprocity and sustainability), but machine learning also enhances editorial workflows.
• Foundation’s Role: The Wikimedia Foundation sees itself as an enabler/facilitator, not a top-down controller, respecting diverse community choices.
• Caution, Collaboration, and Hope: Wikipedia’s ability to adapt, resist toxicity, and remain neutral is tied to its community-centric approach—one that could inspire a healthier, more open future for the Internet at large.

End of summary. For more, visit 404media.co.