Singapore
Singapore regulates AI mainly through voluntary, practical frameworks, and its binding AI-specific rules are limited to deepfakes. Its Model AI Governance Framework, extended to generative AI in 2024 and to agentic AI in January 2026, sets out governance practices, and its AI Verify toolkit lets organizations test systems against those principles. The approach is pro-adoption and interoperability-focused, designed to align with international standards.
Key points
- The Model AI Governance Framework is voluntary and was extended to generative AI in 2024 and to agentic AI in January 2026.
- AI Verify provides a testing toolkit to demonstrate responsible AI practices. On October 14, 2026 IMDA released a starter kit for testing agentic systems, and the AI Verify Foundation launched its AI Tester Accreditation Programme, accrediting five testing firms for one year.
- Singapore emphasizes international interoperability over prescriptive rules.
- Binding deepfake rules ban realistic fakes of election candidates during elections and, since August 17, 2026, make producing AI-generated intimate images without consent an offense. Since June 29, 2026 the Online Safety Commission can issue directions against intimate image abuse under the Online Safety (Relief and Accountability) Act, which covers altered and synthetic images, and where a court claim for intimate image abuse or image-based child abuse is made out, the court awards at least S$5,000 for each image or recording.
- Existing data-protection and sector law continues to apply.
Laws and instruments
Model AI Governance Framework (incl. Generative AI)
A practical governance framework for deploying AI responsibly, with dedicated editions for generative AI (2024) and agentic AI (January 2026, updated May 2026) and the AI Verify testing toolkit.
Elections (Integrity of Online Advertising) (Amendment) Act 2024
Bans publishing online election advertising that realistically depicts a candidate saying or doing something they did not say or do, from the issue of the writ to the close of polling. The Returning Officer can order removal, and individuals and platforms that break the rules face penalties.
Criminal Law (Miscellaneous Amendments) Act 2025 (AI-generated intimate images)
Extends the Penal Code definition of intimate images to wholly AI-generated material and creates an offense of producing intimate images without consent, punishable by up to two years in prison, a fine, or both, with mandatory imprisonment where the person depicted is under 14.
Timeline
- 2024Model AI Governance Framework for Generative AI released.
- January 22, 2025Ban on deepfakes of election candidates during elections takes effect.
- January 22, 2026Model AI Governance Framework for Agentic AI launched.
- June 29, 2026The Online Safety (Relief and Accountability) Act partly commences and the Online Safety Commission opens, covering intimate image abuse and image-based child abuse, deepfakes included.
- August 17, 2026Offense of producing intimate images without consent, including AI-generated images, takes effect.
- October 14, 2026IMDA releases a starter kit for testing agentic systems and forms a Frontier AI Resilience and Security Taskforce, due to report by the first half of 2027; the AI Verify Foundation launches its AI Tester Accreditation Programme with five accredited firms.
Who it applies to
Voluntary for organizations deploying AI in Singapore; binding obligations come from data-protection and sector law and from targeted deepfake offenses.
Penalties
No general AI penalty regime; enforcement is through the Personal Data Protection Act and sector regulators. Deepfake offenses carry fines and prison terms of up to two years.
Common questions
Does Singapore have a comprehensive AI law?
What are the penalties for AI violations in Singapore?
Who does AI regulation in Singapore apply to?
Artificial Intelligence Regulations, "AI Regulation in Singapore," reviewed October 26, 2026, https://artificialintelligenceregulations.com/jurisdictions/singapore.html.
Entries state the position as of the review date and link to the primary text for verification.