National Artificial Intelligence Policy Bill

· About National Artificial Intelligence Policy Bill

Key Takeaways

  • Chile does not yet have a binding AI-specific statute in force; the National AI Policy is a government policy instrument, while the AI bill remains pending in Congress and has not yet created enforceable compliance duties.[2][3][10]
  • The draft AI law would regulate the use of AI systems rather than the technology itself, and it is designed to promote innovation while protecting fundamental rights, democratic principles, and consumer interests.[2][8][10]
  • The bill uses a risk-based model that classifies AI systems into unacceptable, high, limited, and no-evident-risk categories, with obligations increasing with risk level.[8][13]
  • If enacted in its current form, providers and deployers of high-risk AI would face duties such as impact assessments, governance, transparency, documentation, and human oversight, with stronger obligations for systems used in sensitive contexts.[2][8][13]
  • As of October 2026, the bill is still under Senate review and no final enactment date is fixed, so any deadlines are contingent on the law’s passage and any transitional rules the final text may include.[13][15]
  • Current sanctions are not yet available under a standalone AI law, because the bill has not entered into force; once enacted, the text is expected to include administrative enforcement and fines, but the exact penalty scale depends on the final approved version.[10][13]

What It Is

Chile’s “National Artificial Intelligence Policy Bill” refers in practice to the pending bill that regulates AI systems (Boletín 16.821-19, consolidated with 15.869-19), which aims to create a comprehensive framework for AI development, deployment, and oversight in Chile.[2][10][13] The bill is advanced by the Ministry of Science and processed in the National Congress; the Chamber of Deputies approved it and sent it to the Senate for second constitutional review.[2][15]

The policy backdrop is Chile’s updated National AI Policy and Action Plan, which was approved by Decreto 12 of 28 January 2025 and sets government priorities for AI development, ethics, and capability building.[4] That policy is not itself a binding compliance regime, but it informs the legislative approach and the country’s AI governance agenda.[4][8]

Dates and status

  • 7 May 2024: The Executive introduced the AI bill to Congress.[10][13]
  • 13 October 2025: The Chamber of Deputies approved the bill and sent it to the Senate.[13]
  • 2026: The bill remained under Senate discussion and did not become law by October 2026.[13][15]
  • 28 January 2025: Chile approved the updated National AI Policy via Decreto 12.[4]

No official enactment date is available as of 7 October 2026 because the bill remains pending.[13][15]

Who Must Comply

If enacted, the bill would apply to providers, developers, deployers, importers, distributors, and other actors placing AI systems on the Chilean market or using them in Chile, with duties scaling by role and risk category.[8][13] The framework is intended to cover both public- and private-sector uses where AI affects people’s rights, safety, or access to services.[2][10]

The bill is designed with extraterritorial reach in practice for foreign actors whose AI systems are offered in Chile or whose outputs affect persons in Chile, because the legislative purpose is to regulate AI use in the country rather than only domestic companies.[8][13] This means an offshore vendor could be in scope if its system is marketed, integrated, or materially used in Chile.[13]

The available sources do not show finalized statutory exemptions because the law is not yet enacted, but the policy and bill materials indicate that the regime is meant to be proportionate and risk-based, not a blanket prohibition on all AI use.[8][10] General Chilean law would still apply to sectors such as consumer protection, labor, privacy, discrimination, cybersecurity, and product safety even before the AI bill enters into force.[10][13]

Core Requirements

  1. Risk classification: The bill groups AI systems into unacceptable, high, limited, and no-evident-risk categories, and obligations increase with the potential impact on rights and safety.[8][13]
  2. Impact assessments: High-risk uses are expected to require prior or documented AI impact assessment covering foreseeable harms, rights impacts, and mitigation measures.[2][8]
  3. Transparency obligations: Users should be told when they are interacting with AI, and certain synthetic or automated outputs require disclosure or labeling to reduce deception and manipulation.[2][8]
  4. Human oversight: High-risk systems are expected to include meaningful human supervision, escalation paths, and the ability to interrupt or override harmful outputs.[8][13]
  5. Governance and documentation: Providers would need technical documentation, traceability, data governance, and internal controls to show compliance and support regulatory review.[8][13]
  6. Safety and rights protections: The framework seeks to ensure AI respects fundamental rights, democratic principles, and lawful consumer protection, particularly where AI is used in employment, education, public services, or other sensitive contexts.[2][10]
  7. Regulatory supervision: The law contemplates administrative oversight by public authorities, with more intensive review for higher-risk and prohibited uses.[8][13]

Deadlines and Penalties

| milestone | date | what applies | |---|---:|---| | Executive introduction | 7 May 2024 | Bill entered Congress and began legislative processing.[10][13] | | National AI Policy update | 28 January 2025 | Updated policy and action plan approved by decree.[4] | | Chamber approval | 13 October 2025 | Bill passed the Chamber of Deputies and moved to the Senate.[13][15] | | Current status | 7 October 2026 | Bill remains pending in the Senate; no binding AI-law deadlines yet apply.[13][15] |

Because the bill is not yet in force, there is currently no standalone AI-law fine schedule that can be enforced under the pending bill.[13] The final version may include administrative fines, corrective orders, suspension measures, and other sanctions, but the exact maximum penalties cannot be stated definitively until the Senate completes the legislative process and the text is promulgated.[13][15]

How to Comply

  1. Map your AI inventory and classify each system by use case, data inputs, user impact, and likely risk tier, using an approach consistent with ISO 27001 asset inventory and control ownership practices.[8][13]
  2. Build a risk triage method that mirrors the bill’s risk tiers and aligns operationally with NIST CSF 2.0 functions for governance, identification, protection, detection, response, and recovery.[8][13]
  3. Perform impact assessments for sensitive or high-impact use cases, documenting rights impacts, bias risks, safety issues, and mitigations before deployment and after major changes.[2][8]
  4. Implement transparency controls covering user notice, AI disclosure, synthetic content labeling, and customer-facing explanations for material automated decisions.[2][8]
  5. Set human oversight and escalation procedures so a qualified person can review, override, or suspend high-risk outputs, especially in public-facing or consequential decisioning.[8][13]
  6. Strengthen security and governance with ISO 27001-aligned access control, logging, supplier management, incident response, and change control for model and data pipelines.[8]
  7. Adopt AI management-system practices from ISO 42001 to formalize policy, roles, lifecycle controls, and continuous improvement around AI governance.[8]
  8. Track legislative change continuously because compliance obligations may change substantially once the Senate finalizes the bill and transitional periods are set.[13][15]

Related Regulations

  • EU AI Act: Chile’s bill is clearly influenced by the EU’s risk-based structure, but Chile’s text remains pending and may diverge on scope, enforcement, and penalties.[8][13]
  • GDPR / privacy laws: Privacy rules can overlap where AI uses personal data, but the AI bill would address broader AI governance issues beyond data protection alone.[10][13]
  • Chile consumer protection law: Consumer-law duties can already apply to misleading AI outputs or automated commercial practices, even before a standalone AI law takes effect.[10][13]
  • Chile discrimination and labor rules: Existing equality and labor protections can already constrain AI use in hiring, evaluation, and workplace monitoring.[10][13]
  • Cybersecurity regime: Security obligations for systems handling sensitive data or critical services may overlap with AI governance controls, especially for logging, resilience, and incident response.[10][13]

FAQ

Does Chile already have an AI law in force?

No. The available official and secondary sources indicate that Chile’s AI-specific bill is still pending in the Senate and has not yet been promulgated as binding law.[13][15] The country does, however, have an updated National AI Policy approved in 2025, which guides public policy but does not by itself create a compliance regime.[4]

Does the bill apply to companies outside Chile?

Potentially yes, if a foreign provider’s AI system is marketed, deployed, or used in Chile, or produces effects for people in Chile.[8][13] The available materials frame the law around the use of AI systems in Chile rather than purely domestic incorporation status.[10][13]

Will low-risk AI systems face the same obligations as high-risk systems?

No. The bill uses a tiered model in which obligations increase as risk increases, and limited- or no-evident-risk systems are expected to face lighter requirements than high-risk systems.[8][13] That said, transparency and consumer-protection duties may still apply to lower-risk systems in specific contexts.[2][8]

Are AI impact assessments mandatory?

The legislative materials indicate that impact assessments are a core feature for higher-risk uses, especially where AI may affect rights, safety, or access to essential services.[2][8] Because the bill is still pending, the final timing, scope, and trigger thresholds may still change.[13][15]

What penalties could apply once the law is enacted?

The sources available here do not provide a final, enacted sanction table because the bill has not yet become law.[13] A future version is expected to include administrative enforcement tools and fines, but the maximum amounts must be confirmed from the final promulgated text.[13][15]

Sources

Put it into practice

More compliance guides