Radical Transparency: Redefining the Ethics of AI in the Modern Newsroom

By Anika Collier Navaroli

In an era where the news cycle is increasingly accelerated by generative artificial intelligence, the question for journalists is no longer whether we can use these tools, but how we can use them without sacrificing the foundational trust of our readers. As the director of the Craig Newmark Center for Journalism Ethics and Security, I have spent considerable time navigating the legal and ethical minefields of new technology. The most pressing question I receive remains: "What does transparency really mean when it comes to disclosing a journalist’s AI use?"

I argue that we need a more radical approach to transparency. To demonstrate this, I am moving beyond abstract theory to show, rather than tell, how I personally integrate—and reject—AI tools in the creation of this very column.

The Illusion of Fact: A Case Study in Hallucination

Transparency starts with admitting the limitations of the technology we invite into our workflow. During the research phase of my inaugural column, I recalled a story about a tech journalist who had trained a local Large Language Model (LLM) on their own blog archives. I turned to Google’s AI Overview to verify the memory.

The AI confidently corroborated my recollection. Suspicious—as any good journalist should be—I cross-referenced the claim using Gemini, requesting specific citations. Gemini failed to provide a single verifiable source. It became clear that the LLM had simply hallucinated an agreement with my own false memory.

This experience was a stark reminder of the "black box" problem. When an LLM agrees with you, it isn’t necessarily providing truth; it is providing a reflection of your own biases or misconceptions. I ejected the falsehood from my draft immediately. This serves as a cautionary tale: the most "productive" use of AI is often realizing when it is entirely useless.

The Writing Process: Human Alchemy vs. Algorithmic Autocomplete

There is a pervasive fear that AI will eventually replace the journalist, but my own process suggests a different reality. I do not use generative AI to draft my columns, nor do I employ downloadable AI writing assistants.

My approach to prose is deeply human—and, at times, deliberately messy. Inspired by my summer reading of Toni Morrison, I often lean into what Harvard professor Namwali Serpell describes as "breaking grammar." I prioritize the cadence and soul of language over the clinical sterility of automated editing.

However, I am not a Luddite. I utilize the standard features embedded in word processors, such as Google Docs’ spell-check and autocomplete. While I mostly ignore the latter, it occasionally serves a utilitarian function—such as reminding me how to spell "chagrin." These tools are extensions of existing technology, distinct from the generative models that threaten to mimic human creativity.

AI as a "Smarter Thesaurus"

Where generative AI finds its most appropriate niche in my work is as a "smarter thesaurus," a concept popularized by The Atlantic’s Will Oremus. When I am stuck on a specific nuance—such as finding a synonym for "negligence" that starts with "M" or searching for a precise term for "cringe"—ChatGPT serves as a high-speed research assistant.

It is important to draw a bright line here: asking for a synonym is a tool-based interaction; asking for a paragraph is a creative surrender. The former enhances my own vocabulary; the latter abdicates my responsibility as a writer.

The Cringe Factor: Why AI Fails at Social Strategy

Perhaps the most revealing moment in my AI experiment occurred when I attempted to outsource social media promotion. After writing a column on the legal complexities of AI speech rights, I asked ChatGPT to generate a LinkedIn post to promote it.

The result was, in a word, "cringe." It was a hollow, jargon-filled imitation of professional enthusiasm. I closed the tab without using a single word. I concluded that I would rather fail in front of my professional network using my own voice than succeed through the soulless mimicry of an LLM. This highlights a critical limitation of AI: it lacks the context, irony, and genuine human connection that defines effective communication.

Implications for the Industry: A Framework for Disclosure

The goal of this "radical transparency" is to move the industry toward a standardized model of disclosure. Journalists should not be afraid to show their work.

Recommended Disclosure Frameworks:

1. The "How-To" Disclosure: Provide a brief parenthetical note on the specific prompts used and how the output was integrated (e.g., "AI was used to summarize source documents, but all analysis is human-written").
2. Product Naming: Always name the specific AI tools used, whether it is Claude, Gemini, or ChatGPT.
3. Visual Indicators: Newsrooms could adopt icon-based disclosures that link to a detailed policy page, allowing readers to understand the outlet’s stance on AI at a glance.

By adopting these practices, we avoid the "embarrassing after-the-fact disclosures" that plague outlets that hide their AI usage. Transparency isn’t just an ethical duty; it is a mechanism for building institutional legitimacy.

The Ethical Ocean: Why We Must Tread Carefully

We must acknowledge that AI is not a neutral tool. Its development often relies on the unauthorized scraping of intellectual property, and its operational requirements carry a massive climate footprint. There is also the very real danger of job displacement.

Because of these ethical concerns, transparency is our primary defense. If we are going to use these tools, we must be honest about why. When I mentioned to CJR’s editor-in-chief, Betsy Morais, that I always write with a thesaurus, she quipped that she always edits with a dictionary. That simple exchange captured the essence of the editorial process—a collaborative, human-centric friction that no LLM can replicate.

Conclusion: The Alchemy of Human Thought

The "most useful" AI will always be secondary to the "most useful" human. The alchemy between a writer and an editor—the debate over a word choice, the verification of a difficult fact, the insistence on a specific tone—is where the value of journalism resides.

We are currently in a transition period that mirrors the early days of the internet. We can either allow these tools to erode our credibility through opacity, or we can use them to create a new standard of accountability. The choice is ours, but the stakes—the trust of the public—are non-negotiable.

This piece was produced with support from the Craig Newmark Center for Journalism Ethics and Security. Anika Collier Navaroli is an award-winning researcher and lawyer focused on the intersection of AI, law, and journalism.