Complementary Law 205 of 2025, State Policy to Foster Innovation in Artificial Intelligence
Lei Complementar nº 205, de 19 de maio de 2025
In force since .
An AI governance rule binding public and private bodies.
As of .
What it requires
- Give people accessible, free, prior and easily understood information that an AI system is used for automated decisions, what it is for, and that the interaction is automated (article 8)
- When an automated decision directly affects access to or the exercise of rights, or significantly affects interests, give the user the relevant factual premises the system used so the decision can be contested, subject to industrial and commercial secrecy (article 8, paragraph 2)
- Do not discriminate unlawfully through an automated decision; empirically grounded statistical and risk analysis is not unlawful discrimination, but discrimination on grounds of ethnicity, religious belief, sexual orientation or gender stays barred (article 8, paragraph 6)
- Do not develop, implement or use an AI system whose purpose is contrary to fundamental rights, public order, the guiding principles of the democratic rule of law or the security of public institutions (article 9)
- For an autonomous AI agent operating outside the state sandbox, tell end users that it is artificial and decides autonomously, and keep effective human supervision to identify what could harm fundamental rights (article 44, items I and II)
- Have periodic technical audits of the agent's operational security and ethics carried out by independent accredited third parties when it operates as item II provides (article 44, item III)
- For embedded (edge) AI sold or used in public services, tell end users that it exists and what it can and cannot do (article 48, item I)
- Apply reinforced cybersecurity measures to edge AI, including strong encryption, automatic security updates and vulnerability alert systems (article 48, item II)
- Have periodic independent technical audits of edge AI devices used in critical areas such as health, public security, transport and infrastructure (article 48, item III)
- Provide for human supervision and intervention in edge AI, including remote shutdown in critical or emergency situations where technically feasible (article 48, item IV)
- Keep clear, accessible and auditable technical documentation of an edge AI system's operation, operating limits, potential risks and safety (article 48, item V)
- If you deliver a public service under a concession, permission, agreement or partnership with the state, give users of AI-assisted decisions a reasoned decision, the right to contest and ask for review, and human review on appeal, to the extent applicable; entities in free competition or with no formal arrangement with the state are outside this duty (articles 12 and 13)
These lines are LexLint's reading of the law's text, as of .
What this law does
Complementary Law 205 of 2025 institutes the state policy to stimulate the development of innovation in artificial intelligence in Goiás. In automated decisions, people have a right not to be discriminated against unlawfully and to prior, free, accessible and easily understood information about the use of the system, its purpose and the automated character of the interaction.
Article 8 does not apply to systems used solely for security, cyberdefense or military purposes, to systems that take no part in decisions or take none that affect the legal sphere of third parties, or where a human takes the final decision. The Law forbids developing, implementing or using an AI system whose purpose is contrary to fundamental rights, public order, the guiding principles of the democratic rule of law or the security of public institutions.
Outside experimental contexts, autonomous AI agents must meet minimum mandatory requirements: transparency about their artificial nature and autonomous decision-making capacity, effective mandatory human supervision, and, when they operate under that supervision, periodic technical audits by independent accredited third parties.
Developers, operators and distributors of embedded (edge) AI must meet five mandatory requirements: transparency to end users, reinforced cybersecurity, periodic independent audits for devices in critical areas, human supervision and intervention with remote shutdown where technically feasible, and clear technical documentation. Users of AI in state public services have rights to a reasoned decision, to contest and obtain review of it, and to human review on appeal.
Complementary Law 215 of moved the Núcleo de Ética e Inovação em Inteligência Artificial into the administrative structure of the Secretariat of State for Science, Technology and Innovation (SECTI), which now chairs it.
When LexLint raises it
When your app profile says your app deploys a chatbot, distributes a software product or makes high-risk automated decisions.