Pontifícia Universidade Católica do Paraná and Brazilian creative studio 404 Innovation Studio on August 31, 2026 launched a free 12-article protocol that users paste into any AI chat window to change how the system responds, framing the release as a defense of human thinking rather than a product.

The protocol, called the Humanitas Code, works by instructing whichever AI model receives it - ChatGPT, Claude, Gemini or otherwise - to cite sources, present alternative perspectives, and leave final judgment with the person typing rather than the machine answering. It arrives as a block of text that a user copies from a dedicated website and pastes at the start of a conversation, rather than as software installed on a device or a plug-in added to a browser. Five final clauses follow the twelve articles.

PUCPR, a private Catholic research university founded in 1959 in Curitiba, developed the code together with 404 Innovation Studio, a Brazilian creative company under the Galeria.Holding network. The launch was structured as a brand campaign, titled "Never Stop Thinking," with television, radio, cinema, out-of-home and digital placements running across Brazil. A dedicated site, nuncaparedepensar.com.br, hosts the code text along with instructions for copying and sharing it.

What the protocol claims to do

According to the joint announcement, the Humanitas Code went through what the companies describe as a range of evaluation scenarios, including A/B tests conducted with different artificial intelligence models. Those tests, according to the release, aimed to assess whether the protocol could produce consistent changes in the behavior of AI tools across different use cases. Neither the sample size, the specific models tested beyond the three named consumer chatbots, nor the metrics used to judge consistency were disclosed in the material PPC Land reviewed.

The mechanism itself is not novel from a technical standpoint. Pasting standing instructions at the top of a conversation is a known technique often called a system prompt or, when supplied by the user rather than the platform, a custom instruction. The practice works because a language model has no persistent memory between sessions and instead reconstructs its behavior from whatever text sits in its context window at the time it generates a reply. What PUCPR and 404 Innovation Studio have done is package that technique as a public artifact, written for lay users rather than developers, and attach it to institutional and religious framing rather than technical framing.

Cristina Pastore, Director of Brand and Future at PUCPR and the executive credited with conceiving and leading the project, framed the initiative around a distinction between labor displacement and cognitive displacement. "The question we kept coming back to wasn't what AI will do to work; it was what is AI doing to thinking," Pastore said, according to the announcement. "For a university whose mission for nearly 70 years has been to shape people who think, that question isn't academic. The Humanitas Code is our answer. We believe in AI, and we also believe the final decision has always belonged to the human. We want to keep it that way."

Rafael Caldeira, co-founder of 404 Innovation Studio, described the starting point as the volume of commentary already circulating about AI's effects on daily life. "The use of AI is going to change our lives, and we need to give people tools that help them build a better relationship with these technologies," Caldeira said. "The Humanitas Code is a piece of code, from a university, given away to the public, that changes the behavior of the AI in front of you. To our knowledge, it's among the first examples of a brand building cognitive protection directly into an AI tool for the person using it."

The religious anchor

The code's framing borrows directly from Magnifica Humanitas, the first encyclical of Pope Leo XIV's pontificate, which the Pope signed at Saint Peter's on 15 May 2026 and presented publicly at the Vatican on 25 May 2026. PPC Land's coverage of the encyclical detailed a document that called for AI to be "disarmed" from what it terms logics of domination, exclusion and war, and that argued current AI systems "merely imitate certain functions of human intelligence" without undergoing experience, possessing a body or exercising moral conscience. The encyclical called on Catholic educational institutions specifically to preserve in new generations the desire to ask questions, according to PUCPR's own account of the document's instructions.

PUCPR positions itself as responding directly to that call. The university's press material describes the Humanitas Code as an answer to the encyclical's charge, built by an institution whose motto - Scientia, Vita et Fides, or Science, Life and Faith - already frames scientific research inside a religious mission. The convergence of a live papal document on AI ethics and a university-branded consumer product is, on its face, an unusual pairing for the advertising industry to track, but the campaign explicitly seeks the kind of institutional credibility that a Vatican text on artificial intelligence confers.

Ronaldo Lemos, a Brazilian lawyer, professor and technology researcher credited as the lead architect of the Marco Civil da Internet - Brazil's foundational internet rights framework - serves as an advisor to the initiative. Lemos framed the project's significance in practical rather than symbolic terms. "What makes the Humanitas Code particularly interesting, beyond the fact that there are no similar precedents in the world, is that it transforms the responsible use of artificial intelligence into a tool embedded in our everyday interaction with technology," Lemos said, according to the announcement. "Rather than simply discussing the future of AI, we have created mechanisms to influence this relationship in the present."

Paulo Aguiar, a Brazilian educator and creative consultant specializing in generative AI who also contributed to the protocol's development, added a similar point about application over commentary. "It is precisely this practical application that makes the Code relevant: transforming principles of responsibility and ethics into guidelines for the choices and interactions we make with AI every day," Aguiar said, according to the announcement.

Lemos has also published a book addressing related territory. O Monge, o Iogue e o Faquir - Corpo, Coração e Mente no Tempo das Máquinas ("The Monk, the Yogi, and the Fakir - Body, Heart, and Mind in the Age of Machines"), released by Brazilian publishing house Todavia, takes up what the announcement describes as the same central thesis as the Humanitas Code: how to preserve what is human as machines take on more of daily cognitive work.

Why the timing matters for a marketing audience

The mechanism behind the Humanitas Code sits squarely inside a debate that PPC Land has tracked closely over the past year: whether frequent, unstructured delegation of reasoning to AI systems measurably degrades the cognitive skills involved in evaluating information. A February 2025 study from Microsoft Research and Carnegie Mellon University, covered by PPC Land, surveyed 319 professionals who regularly use generative AI tools and found that 83 of them reported that trust and reliance on AI tools discouraged critical reflection on their own work. The same research found that workers with higher confidence in AI capabilities exerted less critical thinking effort, while those confident in their own abilities engaged in deeper analysis despite perceiving it as more effortful.

That dynamic recurred in a separate report PPC Land covered in February 2026, which examined the gap between how workers describe using AI for writing tasks and what usage data actually shows. The report cited a 2024 study linking overreliance on large language models among students to measurable weakening of critical thinking, decision-making and analytical abilities. A more visible instance of the same pattern surfaced in June 2026, when PPC Land reported on a Florida appellate hearing in which a licensed attorney admitted under questioning from three judges that the argument section of his brief, generated with AI assistance, contained no argument of any kind. The 44-minute proceeding drew nearly 8,000 views within days of its upload and circulated widely across professional networks, feeding a broader debate about what happens to specialized knowledge when tools perform the reasoning step rather than merely the drafting step.

For an industry that has spent the better part of two years wiring AI systems directly into media buying, the Humanitas Code lands in an environment already saturated with related but distinct concerns: content provenance, disclosure obligations and system transparency. The EU AI Act's Article 50 transparency obligations, which PPC Land has covered extensively since the European Commission finalized its guidelines and Code of Practice on 20 July 2026, became legally applicable on 2 August 2026 - twenty-nine days before the Humanitas Code launched. That regulation compels providers and deployers of generative AI systems operating in the European Union to disclose AI-generated content and mark synthetic material, addressing what the output of a system looks like. The Humanitas Code addresses something different and upstream of that: what instructions shape the system's behavior before it produces any output at all, and whether a user retains enough independent judgment to evaluate what comes back.

The distinction matters because the two efforts are not substitutes. Article 50 is statutory, backed by fines that PPC Land reported can reach 3 percent of global turnover for non-compliance, and it binds companies regardless of whether an individual user takes any action. The Humanitas Code is voluntary at every level: no platform is obligated to honor it, no regulator enforces it, and its effect depends entirely on whether a person chooses to paste roughly a page of text into a chat window before asking a question. Whether AI providers' underlying systems reliably follow instructions pasted by end users - as opposed to instructions set by the platform itself in its own system prompt - is an open technical question the announcement does not address, and one that varies by model and by how a given platform prioritizes competing instructions.

Enforcement problem and precedent

The protocol's central mechanical claim, that a block of pasted text changes how a large language model behaves, has some technical grounding. A prompt - the term PPC Land's own explainer series defines as everything a language model receives before producing output - is the primary lever available to end users who cannot retrain or fine-tune the underlying model. Commercial chat interfaces from OpenAI, Anthropic and Google all support user-supplied instructions that persist for the length of a conversation or, in some products, across sessions if saved as a standing preference. Whether those instructions survive contact with a platform's own system-level rules, safety guardrails or built-in behavioral defaults is not guaranteed, and the announcement offers no technical documentation on how the Humanitas Code's twelve articles interact with those underlying layers when they conflict.

The campaign describes itself as, to the companies' knowledge, among the first examples of a brand embedding what it calls cognitive protection directly into an AI tool for the end user, as distinct from AI governance work aimed at companies, regulators or platforms. That claim of novelty is difficult for PPC Land to independently verify, since the broader landscape of AI governance work has concentrated overwhelmingly on provider-side and regulator-side interventions rather than user-side prompt templates distributed as brand marketing.

PUCPR brings substantial institutional scale to the campaign regardless of the protocol's technical reach. The university enrolls more than 54,000 students and employs roughly 1,800 researchers across three campuses in Paraná state - Curitiba, Toledo and Londrina - plus an international hub in Florence, Italy, spanning 17 master's and doctoral programs across six schools. Independent rankings cited in the announcement, Times Higher Education 2025 and Ranking Universitário Folha 2025, both placed PUCPR third among private universities in Brazil, and Ranking Universitário Folha separately named it the country's most innovative private university.

404 Innovation Studio, founded in 2023 and formerly known as 404 Design & Innovation, operates as one of thirteen companies inside Galeria.Holding, an independent Latin American marketing and communications group with roughly 900 employees led by chairman Eduardo Simon. The studio describes its own approach as design-led and collaborative, built around solutions for what the announcement calls transformative brands.

What remains unverified

Several claims in the announcement fall outside what PPC Land can independently confirm from the material provided. The specific AI models used in the A/B evaluation phase, the number of test conversations run, and the criteria used to judge whether the protocol produced consistent behavioral change are not disclosed. Nor does the announcement specify how the code performs when a platform's own system instructions directly contradict an article of the Humanitas Code, a scenario that determines whether the protocol functions as advertised in adversarial conditions rather than cooperative ones. The claim that the initiative is "among the first" of its kind is presented by the companies themselves rather than substantiated against a documented landscape of comparable prior efforts.

Timeline

Summary

Who: Pontifícia Universidade Católica do Paraná (PUCPR), a Catholic research university in Curitiba, Brazil, together with 404 Innovation Studio, a Brazilian creative company within the Galeria.Holding network. Cristina Pastore, PUCPR's Director of Brand and Future, led the project; Rafael Caldeira co-founded 404 Innovation Studio; Ronaldo Lemos and Paulo Aguiar advised on development.

What: The launch of the Humanitas Code, a free 12-article protocol plus five final clauses that users paste into any AI chat interface to instruct the system to cite sources, present alternative perspectives and preserve final judgment with the user, alongside a brand campaign called "Never Stop Thinking."

When: The protocol and campaign launched on 31 August 2026, roughly three and a half months after Pope Leo XIV signed the encyclical Magnifica Humanitas on 15 May 2026 and four weeks after EU AI Act Article 50 transparency obligations took legal effect on 2 August 2026.

Where: Developed in São Paulo and Curitiba, Brazil, and distributed globally through a dedicated website, with the initial marketing campaign running across Brazilian television, radio, cinema, out-of-home and digital channels.

Why: The protocol responds to Magnifica Humanitas, which called on Catholic educational institutions to preserve younger generations' capacity to ask questions rather than defer wholesale to AI systems. It arrives amid mounting research - including studies PPC Land has covered on workplace critical thinking, AI-assisted writing and a widely viewed courtroom failure - documenting how habitual reliance on AI systems can weaken independent analysis, judgment and verification skills.