CNPq Integrity Policy: Ordinance 2664/2026

CNPq (Conselho Nacional de Desenvolvimento Científico e Tecnológico — the National Council for Scientific and Technological Development) is Brazil’s principal federal science-funding agency, linked to the Ministry of Science, Technology and Innovation (MCTI). On March 6, 2026, CNPq published Portaria (Ordinance) nº 2.664/2026 in the Diário Oficial da União, formally instituting a Política de Integridade na Atividade Científica (Policy of Integrity in Scientific Activity). The ordinance replaces the agency’s prior conduct code (in force since August 2024) and is CNPq’s current governing framework for research integrity, research misconduct, and—for the first time in a Brazilian federal science agency’s formal policy—rules on the use of generative artificial intelligence in funded research.

This guide summarizes what the ordinance covers, how CNPq defines and handles research misconduct, how its integrity committee process works when an allegation is raised, and what CNPq-funded researchers and their host institutions are expected to do under it. It draws directly on CNPq’s own published notice and the ordinance text; where a provision could not be independently confirmed, that is noted rather than stated as fact.

CNPq’s Role, in Brief

CNPq is one of Brazil’s two largest general-purpose federal science funders (alongside CAPES, which is oriented specifically toward graduate education and scholarships—see CASRAI’s CAPES guide). CNPq funds research fellowships, productivity grants, and institutional research programs across all fields, and it also operates the Currículo Lattes, Brazil’s standard national researcher-CV platform used to evaluate applicants and track scientific output. For background on CNPq’s funding lines and how international researchers typically engage with the agency, see CASRAI’s separate CNPq funding overview — this guide covers a different, narrower topic: the integrity policy and misconduct-handling process specifically, not CNPq’s grant programs.

Ordinance 2664/2026: Scope and Core Principles

The policy applies broadly—not only to CNPq’s own staff, but to proponentes, beneficiários e demais agentes vinculados ao fomento (proposers, grant beneficiaries, and other agents linked to CNPq funding), as well as anyone using CNPq’s digital platforms and databases, including the Lattes system. In practice, this means any researcher who holds, has applied for, or is named on a CNPq-funded project or fellowship falls within its scope, regardless of institutional affiliation or nationality.

The ordinance is organized around a set of foundational principles rather than a single narrow rule, including: intellectual honesty and good scientific practice; accurate attribution of authorship and scientific credit; respect for research participants and their safety; responsible supervision of students and early-career researchers; compliance with applicable law and institutional policy; and equity, diversity, and civility in research relationships. These principles frame everything that follows—the misconduct definitions and sanctions exist to protect them, not the reverse.

How CNPq Defines Research Misconduct

The ordinance addresses the same core misconduct categories recognized internationally—fabrication, falsification, and plagiarism—alongside conduct specific to CNPq’s own systems and processes:

Fabrication and falsification of data, results, or methods used in CNPq-funded work or in materials submitted to CNPq.
Plagiarism, including self-plagiarism (recycling one’s own previously published material as if it were new, without disclosure).
Currículo Lattes inconsistencies — false, inflated, or misleading entries in a researcher’s Lattes CV, which CNPq treats as a distinct, examinable integrity concern given how heavily the platform is used in funding and evaluation decisions.
Authorship misattribution and failures of responsible supervision.
Discriminatory conduct within the research environment, addressed under the same integrity framework rather than as a separate policy.

For CASRAI’s general definitions of these terms as used across the research-integrity literature, see the dictionary entries for fabrication, falsification, plagiarism, and research misconduct generally.

Generative AI: Declared Use, Not a Blanket Prohibition

Ordinance 2664/2026 is notable for addressing generative AI explicitly—among the first formal policies from a Brazilian federal science agency to do so. It does not prohibit AI tools outright. Instead it requires:

Mandatory declaration of any generative AI use, identifying the tool and how it was used.
No AI authorship — AI-generated content cannot be submitted or presented as if it were the researcher’s own original work without disclosure.
Full researcher responsibility for the final content, including any plagiarism, fabricated citations, or factual errors an AI tool introduces — the human author, not the tool, is accountable.
AI use in peer review is discouraged, reflecting confidentiality and independent-judgment concerns that apply to review processes generally.

This mirrors a broader convergence across research-integrity bodies internationally (COPE and ICMJE have issued comparable positions: disclosure required, no AI authorship, human accountability for accuracy) — CNPq’s approach is consistent with, not a departure from, that emerging consensus.

The Research-Integrity Committee Process (CIAC)

Allegations and cases under the policy are handled by CNPq’s Comissão de Integridade na Atividade Científica (CIAC) — the Scientific Activity Integrity Commission. As described in CNPq’s own published guidance:

Concerns can be raised through CNPq’s Ombudsman (Ouvidoria) channel.
CIAC examines specific concerns referred to it, including suspected Lattes CV inconsistencies and allegations of fabrication, falsification, plagiarism, or other violations of the policy’s core articles.
CIAC is the body that deliberates on findings and, where warranted, on the sanction to apply, using the graduated framework described below.

Importantly, CIAC does not operate as the sole investigative body for every case. Brazilian host institutions (ICTs, or Instituições Científicas, Tecnológicas e de Inovação) carry a parallel, often first-line responsibility described below, with CIAC acting on the CNPq-specific consequences (funding status, eligibility for future selection processes, restitution) once an allegation reaches it.

Role of Host Institutions (ICTs)

The ordinance treats research integrity as a shared responsibility between CNPq and the institution hosting the funded research. ICTs are expected to:

Conduct preliminary investigation of misconduct allegations arising within their own institution.
Cooperate with CNPq audits and share investigation outcomes and any corrective or disciplinary measures taken.
Maintain their own institutional culture and mechanisms for preventing and addressing integrity violations, consistent with CNPq’s framework.

This dual-track structure — institutional first-line handling plus a CNPq-level commission for funding-specific consequences — is a common pattern across national funders internationally; it parallels, for example, how U.S. federal agencies rely on institutional research-integrity officers before ORI or a funding agency’s own office becomes involved. For the general concept, see CASRAI’s research integrity entry.

Sanctions and Consequences

The ordinance sets out a graduated, proportionate sanctions framework rather than a single penalty for any violation of its core articles. Sanctions reported in CNPq’s own summary of the policy include, in increasing severity:

Formal warning.
Temporary suspension of a grant or fellowship.
Termination of the benefit, potentially with restitution of funds already disbursed.
Time-limited barring from applying to or participating in future CNPq selection processes.
Revocation of an award obtained through false qualification claims, with proportional restitution.

In determining severity (“dosimetry”), CNPq’s policy directs the commission to weigh the seriousness and scope of harm caused, evidence of intent or fraud, and any prior recidivism, alongside aggravating or mitigating circumstances. Notably, the policy explicitly excludes a researcher’s academic rank, seniority, or prior awards and recognition from being treated as a mitigating factor — a deliberate signal that institutional stature does not buy leniency under the framework.

What This Means for CNPq-Funded Researchers and Collaborators

For researchers holding CNPq funding, or international collaborators on a CNPq-funded project, the practical takeaways are:

Keep your Currículo Lattes entries accurate and current — discrepancies are an explicitly examinable integrity concern under this policy, not just an administrative nuisance.
Disclose any generative AI use in proposals, reports, or manuscripts tied to CNPq funding, and verify AI-assisted output yourself before it goes out under your name.
Understand that an allegation on CNPq-funded work may be investigated at two levels — your host institution first, and CIAC for CNPq-specific consequences — and that cooperation is expected from both the researcher and the institution.
Review your host institution’s own research-integrity procedures alongside CNPq’s policy; the ordinance sets the federal-funder framework, but institutional (ICT) policy governs day-to-day process at your own organization.

Frequently Asked Questions

What is CNPq’s Scientific Integrity Policy?

It is the framework, formally instituted by Portaria (Ordinance) 2664/2026 on March 6, 2026, that defines research-integrity principles, misconduct categories, generative-AI-use rules, and sanctions for anyone connected to CNPq funding, fellowships, or digital platforms — including the Currículo Lattes system.

Who does the policy apply to?

CNPq staff, grant proposers and beneficiaries, and any other agent linked to CNPq funding or its digital services and platforms — in effect, anyone who holds or applies for CNPq support, or who uses the Lattes CV system, regardless of nationality or institutional location.

How is a research-integrity investigation triggered at CNPq?

Concerns can be raised through CNPq’s Ombudsman channel. From there, cases involving suspected misconduct or Lattes CV inconsistencies can be referred to the Comissão de Integridade na Atividade Científica (CIAC), CNPq’s dedicated integrity commission, while the researcher’s host institution typically conducts a parallel preliminary investigation.

Does the policy ban the use of AI in research proposals or manuscripts?

No. It requires declared, transparent use of generative AI tools and holds the human author fully responsible for the final content, but it does not prohibit AI assistance outright. It does not recommend using AI in peer review.

What happens if misconduct is confirmed?

Consequences range from a formal warning through suspension or termination of funding, restitution of disbursed funds, and time-limited exclusion from future CNPq selection processes, depending on the severity, intent, and scope of harm involved.

Related CASRAI Resources

Sources: CNPq, “CNPq institui Política de Integridade na Atividade Científica” (gov.br, March 2026); Portaria CNPq nº 2.664, de 6 de março de 2026, published in the Diário Oficial da União. This guide summarizes CNPq’s own published description of the ordinance; readers relying on specific provisions for compliance purposes should consult the full ordinance text and their institution’s research-integrity office.