Every company that processes personal data or depends on technology needs policies: on privacy, on information security, on the use of IT resources and, increasingly, on the use of artificial intelligence. Many meet the requirement the fastest way possible: a ready-made template with the company's name swapped in. It looks like a cheap solution. In this article, we show why it often turns out to be expensive.

A generic policy does not cover the risks of the business or the regulatory requirements that actually apply to that company. Because it is a document that can be read, discussed and used in administrative and judicial proceedings, a text that does not match practice stops protecting the company and starts documenting the gap.

On video (in Portuguese): Guilherme Goulart explains why a generic policy does not protect the company (3 min).

What is the problem with using a ready-made policy template?

A policy is the simplest way to demonstrate compliance to the outside world: customers, partners, investors and the authority. The dynamic is well known in law. It is the same as that of the generic contract, used to regulate private interests it knows nothing about: it fits any business and, for that very reason, fits none of them well.

There is an important difference, however. A policy is not kept in a drawer. It is a document that can be read, discussed and used in administrative and judicial proceedings. When it is generic, it does not cover the risks of the business or the regulatory requirements that actually apply to that company.

What does the LGPD require of data protection policies?

The LGPD, Brazil's General Data Protection Law (Law No. 13,709/2018, in Portuguese), does not require every company to adopt rules of good practice and governance: Art. 50 says that controllers and processors "may" draw them up. But the law does require other things. Processing agents must adopt technical and administrative security measures (Art. 46), and the principle of accountability requires demonstrating effective measures and the effectiveness of those measures (Art. 6, X).

When dealing with good practices, the law requires taking into account the nature, scope, purpose, and the likelihood and severity of risks (Art. 50, § 1), as well as the structure, scale and volume of operations and the sensitivity of the data processed (Art. 50, § 2). None of these elements comes in a template. As Menke and Goulart (2021) put it: "One can only prevent what is foreseen" (free translation).

Nor is a policy a decorative piece. Souza (2019) notes that the information security policy "works as a code of conduct to be followed by employees" (free translation). It was made to be complied with. That is why Carvalho, Mattiuzzo and Ponce (2021) warn that "Implementing a good practices policy is far from being a list of checkboxes that each company fills out without further reflection" (free translation). The authors point out that individually designed programs "have the advantage of being tailor-made" and that even sector rules require "an individualized assessment by the controller or processor of the risks actually faced by the entity on a daily basis" (free translation).

How can a generic policy be used against the company?

The most underestimated risk is the policy being used against whoever published it. In consumer relations, Brazil's Consumer Protection Code (Law No. 8,078/1990, in Portuguese) establishes that sufficiently precise information binds the supplier and becomes part of the contract (Art. 30). If the policy promises what the company does not do, the promise can be enforced.

In the regulatory sphere, the LGPD provides for the controller to demonstrate the effectiveness of its governance program, in particular at the request of the national authority (Art. 50, § 2, II). In court proceedings, the judge may shift the burden of proof in favor of the data subject (Art. 42, § 2). In these scenarios, a text that does not match practice stops protecting the company and starts documenting the gap.

There is an irony here. The law includes "the adoption of a policy of good practices and governance" among the criteria for applying sanctions (Art. 52, § 1, IX), a parameter reproduced in the ANPD's dosimetry regulation, CD/ANPD Resolution No. 4/2023 (OLIVEIRA, 2024). A real policy can weigh in the company's favor. A façade policy carries no weight and exposes the company as well.

And it is not only the authority that reads it. In negotiations, customers can use the generic document to question the company's maturity, and competitors can use it to try to discredit its security and data protection practices. It is a clear, and avoidable, demonstration that the company does not take the subject seriously.

Can artificial intelligence write the company's policy?

With AI, generating a policy has become a matter of seconds. But the problem was never the writing. As Marcolini (2024) observes, "The assessment of the appropriate level of security must be made by the controller in the specific case, always through an analysis of the risk to the protection of personal data" (free translation).

Many requirements depend on each company's situation: its contracts, its relationships, the types of data it processes, its market position and, above all, a well-structured risk analysis. AI can help with the writing. It does not make that assessment on the company's behalf.

Nor is the argument that a generic policy is better than none convincing. Luz (2024) shows that "the analysis of compliance programs has evolved to address the following question: what should be considered an effective program" (free translation). Having the document is no longer enough.

Where to start building the company's policy

  1. Take inventory: what data the company processes, where, for what purpose and with whom it is shared.
  2. Carry out the risk analysis before writing the policy, not after.
  3. Read the contracts with customers and suppliers: many security obligations are in them.
  4. Write only what the company complies with, and comply with what is written.
  5. Define who is responsible and how often the policy is reviewed (the LGPD provides for periodic updating of good practice rules in Art. 50, § 3).
  6. Keep evidence of implementation: training, records, reviews.
  7. Use AI as a writing aid, never as a substitute for the assessment.

BrownPipe Consultoria helps companies build technology policies based on risk analysis and the reality of each business.

Frequently asked questions

Does the LGPD require companies to have a policy of good practices and governance?

No. Art. 50 of the LGPD says that controllers and processors "may" draw up rules of good practice and governance. The law does, however, require the adoption of technical and administrative security measures (Art. 46) and the demonstration that these measures are effective (Art. 6, X).

Is using a ready-made policy template better than having none?

The argument does not hold up. A generic policy does not cover the risks of the business or the applicable regulatory requirements, and the analysis of compliance programs now asks whether the program is effective, not whether the document exists.

Can a privacy or security policy be used against the company?

Yes. In consumer relations, sufficiently precise information binds the supplier and becomes part of the contract (Consumer Protection Code, Art. 30). The LGPD provides for the controller to demonstrate the effectiveness of the governance program at the request of the ANPD (Art. 50, § 2, II), and the judge may shift the burden of proof in favor of the data subject (Art. 42, § 2).

Does having a good practices policy help in an ANPD sanction?

It can weigh in favor. The LGPD includes the adoption of a policy of good practices and governance among the criteria for applying sanctions (Art. 52, § 1, IX), a parameter reproduced in CD/ANPD Resolution No. 4/2023. A façade policy has no such effect and also exposes the company.

Can artificial intelligence write the company's security policy?

It can help with the writing, but it does not replace the assessment. Many requirements depend on contracts, relationships, the types of data processed and, above all, a risk analysis carried out in the specific case.

References

CARVALHO, Vinicius Marques de; MATTIUZZO, Marcela; PONCE, Paula Pedigoni. Boas práticas e governança na LGPD. In: DONEDA, Danilo et al. (coord.). Tratado de proteção de dados pessoais. Rio de Janeiro: Forense, 2021.

LUZ, Ilana Martins. Elementos para um programa de compliance eficaz. Revista de Direito Penal Econômico e Compliance, v. 17, p. 71-98, jan./mar. 2024.

MARCOLINI, Bruno Junqueira Meirelles. Entre riscos e obrigações: a atuação do controlador em incidentes de segurança. In: AGÊNCIA NACIONAL DE PROTEÇÃO DE DADOS. II Prêmio Danilo Doneda de Artigos Científicos: artigos científicos vencedores. Brasília: ANPD, 2024.

MENKE, Fabiano; GOULART, Guilherme Damasio. Segurança da informação e vazamento de dados. In: DONEDA, Danilo et al. (coord.). Tratado de proteção de dados pessoais. Rio de Janeiro: Forense, 2021.

OLIVEIRA, Kátia Adriana Cardoso de. Formação de jurisprudência administrativa pela ANPD: estudo de casos das sanções aplicadas. Revista Digital de Direito Administrativo, Ribeirão Preto, v. 11, n. 2, 2024.

SOUZA, Carlos Affonso Pereira de. Segurança e sigilo dos dados pessoais: primeiras impressões à luz da Lei 13.709/2018. In: FRAZÃO, Ana; TEPEDINO, Gustavo; OLIVA, Milena Donato (coord.). Lei Geral de Proteção de Dados Pessoais e suas repercussões no direito brasileiro. São Paulo: Thomson Reuters Brasil, 2019.