In a 2026 legal proceeding, Mariana Zuvic’s inability to answer Aníbal Ibarra’s questions underscores AI’s fraught role in journalism, exposing gaps in accountability, data ethics and technical transparency.
The Ethical Quandary of AI-Driven Journalism
The Causa Cuadernos case crystallizes a systemic crisis: when AI systems curate or generate news, who bears responsibility for inaccuracies? Zuvic’s evasiveness during Ibarra’s interrogation hints at a broader issue—the opacity of algorithms that now shape public discourse. The video “¿Puede sobrevivir la inteligencia artificial sin periodismo?” frames this as a philosophical debate, but the technical reality is more urgent. AI models, trained on fragmented datasets, often replicate biases or omit context, yet their “objectivity” is rarely scrutinized.
Consider the LLM-300B architecture, a 300-billion-parameter model used by several news platforms. Its training data includes 2020–2025 web crawls, but arstechnica.com reveals 42% of its corpus consists of unverified user-generated content, creating a “noise-to-signal” ratio that undermines journalistic rigor. This isn’t hypothetical—GitHub hosts open-source tools to audit such models, but adoption remains sparse.
The 30-Second Verdict
AI’s integration into journalism risks eroding trust unless technical accountability mechanisms are mandated.
Technical Deep Dive: Model Architecture and Training Data
AI systems like NewsForge 2.0 employ transformer-based architectures with end-to-end encryption for data pipelines, yet their decision-making remains a “black box.” A 2026 IEEE study found that 68% of AI-generated news articles lacked traceable sources, a direct consequence of LLM parameter scaling that prioritizes fluency over fidelity.
Training data ethics are equally problematic. The La Caja de Victoria video highlights how AI tools often scrape content from paywalled archives, violating licensing agreements. This isn’t just a legal issue—it’s a technical one. RFC 9200 outlines data attribution standards, but compliance is voluntary. As Dr. Lena Choi, CTO of OpenNews, notes: “Without mandatory data provenance, AI becomes a vector for intellectual property theft and misinformation.”
What So for Enterprise IT
News organizations adopting AI must balance speed with compliance, a tension exacerbated by proprietary platforms’ closed ecosystems.
Ecosystem Implications: Open Source vs. Proprietary Tools
The tech war over AI journalism mirrors broader battles between open-source and closed ecosystems. Platforms like NewsForge lock users into proprietary APIs, while projects like HuggingFace offer open-source models with fine-tuning capabilities. This divide isn’t just philosophical—it’s economic. Gartner reports that 73% of newsrooms using closed AI tools face vendor lock-in, limiting their ability to audit or modify systems.
Open-source alternatives, however, aren’t without flaws. DeeplNews, an open-source AI news aggregator, relies on community contributions, leading to inconsistent quality. As Joel Ramirez, cybersecurity analyst at CyberShield
“Open-source tools demand technical expertise to secure. A 2025 breach at a major outlet showed how unpatched dependencies can expose sensitive data.”
The 30-Second Verdict
Proprietary AI tools prioritize profit over transparency, while open-source options require resources to maintain securely.
Expert Perspectives on AI Transparency
The Causa Cuadernos case
