The Ghost in the Newsroom: Reevaluating Authorship in the Age of Generative AI

By Raphael ‘Tsavkko’ Garcia

When Folha de S.Paulo, Brazil’s most influential newspaper, acknowledged in February 2026 that columnist Natalia Beauty had relied extensively on generative artificial intelligence (AI) to produce her opinion pieces, the ensuing firestorm transcended the boundaries of a single newsroom. What began as a local controversy over editorial standards quickly morphed into a global case study on the nature of truth, the definition of authorship, and the future of the Fourth Estate.

As newsrooms worldwide grapple with the integration of Large Language Models (LLMs), the Folha incident serves as a pivot point. It forces a reckoning with a fundamental question: In an era where machines can simulate the human voice with startling precision, what does it mean to "write" an opinion?

The Anatomy of the Controversy

The immediate, visceral reaction to the news—that a prominent voice in Brazilian media was outsourcing her intellectual output to an algorithm—centered on a simple, albeit flawed, question: How much AI is too much?

However, as the dust settled, the inquiry deepened. If an opinion column, which is inherently predicated on a unique, human worldview, can be drafted by a machine, does the byline still hold weight? Is authorship defined by the mechanical act of typing, the conceptual heavy lifting of argument construction, or merely the final act of signing one’s name and accepting liability?

This debate has polarized Brazilian journalists, academics, and editors. On one side are the techno-pragmatists, who view AI as the latest in a long lineage of writing tools—no different, they argue, than the transition from quill to typewriter or typewriter to word processor. On the other side are the traditionalists, who contend that opinion journalism occupies a sacred space where the reader expects a distinctly human consciousness, one shaped by lived experience rather than statistical probability.

A Chronology of the AI Shift

The timeline of this disruption is accelerating. For years, AI was relegated to the periphery of newsrooms—used for data scraping, weather reports, or basic financial summaries. The shift in 2026, punctuated by the Folha scandal, marked a migration of AI into the core of the intellectual product: the opinion column.

• Pre-2026: AI is largely viewed as a productivity tool for mundane, repetitive tasks.
• February 2026: Folha de S.Paulo acknowledges the reliance of Natalia Beauty on generative tools, triggering a national debate on disclosure and intellectual integrity.
• August 2026: Anthropic announces that its new Claude models will embed invisible, machine-readable watermarks in generated text, aligning with the EU AI Act’s transparency requirements.
• Late 2026: The industry reaches a consensus that the "watermarking" of AI text, while technically feasible, does not solve the moral dilemma of authorship.

The introduction of watermarking technology is a significant development. By embedding an invisible, machine-readable signature into text, companies like Anthropic are attempting to create a "provenance of prose." Yet, as critics point out, a watermark only proves the origin of the text; it does not adjudicate the intent. It can tell us that a machine wrote the words, but it cannot tell us if the human editor used that machine to polish a brilliant idea or to substitute for the absence of one.

The Philosophical Crisis of Authorship

Veteran journalist and writer Paulo Markun suggests that the obsession with "how much" the machine wrote is a red herring.

"I wouldn’t draw the line based on how much the machine wrote, but on another question: who decided what the text wants to say?" Markun argued in an interview. "If AI helps organize, test, summarize, or challenge the author’s arguments, we are in the realm of assistance. If it starts determining the thesis, the criteria of relevance, the hierarchy of facts, and the point of view, then it is no longer assistance; it becomes a substitution of the author’s intellectual contribution."

Markun points out that writing has never been a solitary act. Journalism has always relied on the collective labor of editors, researchers, and fact-checkers. However, he draws a sharp distinction between "authorship" and "curation."

"The author defines the relevant question, chooses the angle, assumes the risk of interpretation, and decides what enters and what stays out," Markun says. "The curator merely selects, organizes, and validates something produced by someone else."

The "Floor" vs. The "Ceiling"

Technology journalist Cora Rónai views the Folha case as a symptom of a broader, collective unpreparedness. For Rónai, the most uncomfortable truth revealed by the rise of AI is that, in many cases, it performs at a level superior to average human output.

"AI has raised the floor of writing," Rónai observes. "It ensures that the general standard of prose is higher, but it hasn’t necessarily raised the ceiling. It can produce competent, grammatically perfect, and logical text, but it rarely produces the ‘exceptional.’"

This creates a new challenge for the profession. If "competent prose" is now a commodity produced by an algorithm, the professional value of the journalist must pivot toward judgment, nuance, and the "human signature"—the very things that LLMs, by design, struggle to replicate consistently.

Implications for Credibility and Regulation

Erick Felinto, a professor of media studies at Rio de Janeiro State University (UERJ), warns that the stakes are higher for journalism than for other industries.

"Traditional journalism is already struggling with a massive crisis of credibility," Felinto notes. "This crisis tends to deepen if readers discover that an opinion article—the one place they look for human perspective—was largely generated by a machine."

The Regulatory Landscape: The EU AI Act

The industry is beginning to see the arrival of external guardrails. As of August 2026, Article 50 of the EU AI Act has become the gold standard for transparency. It mandates disclosure when AI is used to generate content on matters of public interest. Crucially, the regulation does not ban AI; it demands accountability. If an article is AI-assisted, the human editor must now stand as a legal and editorial guarantor.

This mirrors the emerging consensus among industry leaders: the issue is not the use of AI, but the secrecy surrounding its use.

The Path Forward: What Defines a Journalist?

As the dust settles on the Folha de S.Paulo case, the definition of a journalist is being fundamentally rewritten. If the future of opinion journalism depends less on the mechanical labor of writing and more on the intellectual ownership of the argument, newsrooms must adapt in three ways:

1. Mandatory Disclosure: If a machine contributes to the structural or narrative development of a piece, the reader deserves to know. Transparency is the only currency that can buy back lost trust.
2. Intellectual Leadership: Editors must prioritize columnists who bring original research, distinct life experience, and a willingness to challenge dominant narratives—qualities that go beyond the capabilities of prompt-based generation.
3. The "Human" Value-Add: As AI handles the synthesis of information, the journalist’s role must evolve into that of an interpreter. The capacity to doubt one’s own inferences, to resist the pressure of the majority, and to stand behind a controversial take is the new benchmark for professional survival.

In the end, the Natalia Beauty controversy may be remembered as the moment the industry stopped asking if machines could write like humans and started asking what, exactly, we still need humans for. The answer is not in the syntax, but in the accountability. A machine can draft an opinion, but it cannot stand in a courtroom or a town square to defend it. As long as that remains true, the human journalist still has a future—provided they are willing to own the thinking, not just the text.