Deepfake laws around the world

Singapore's OSRA 2025 and pending Criminal Law Bill create an Online Safety Commission and extend intimate-image offences to AI-generated NCII.

Singapore's 2024 Act bans deepfakes of election candidates during campaign periods, with platform fines up to S$1 million and Returning Officer takedowns.

India's 2025 IT Rules amendment requires labeling and takedown of synthetically generated content, with deepfake provisions in force from 20 February 2026.

Australia's 2024 Criminal Code amendment created offences for transmitting non-consensual sexually explicit material, including AI-generated deepfakes.

Asia's first comprehensive AI law sets safety and transparency duties for high-impact AI and marking rules for generative output, phasing in through 2026.

South Korea's 2024 amendments criminalize producing, sharing, possessing, and viewing sexually explicit deepfakes, dropping the intent-to-distribute test.

China's 2025 labeling Measures require explicit and machine-readable labels on AI-generated text, image, audio, and video across Chinese platforms.

China was first to regulate generative AI, requiring real-name verification, security assessments, content review, and watermarking for synthetic media.

Brazil's AI Bill 2338/2023 sets a risk-based framework awaiting the Chamber of Deputies, while TSE Resolution 23.732/2024 already governs election deepfakes.

Canada's AIDA would have set risk-based duties on high-impact AI, but Bill C-27 died at prorogation in January 2025, with a replacement expected in 2026.

The Colorado AI Act is the first US state horizontal AI law, requiring impact assessments and consumer notice for high-risk systems from 30 June 2026.

California's AI package spans frontier transparency, provenance, election deepfakes, and companion chatbots, led by SB 53, SB 942, and SB 243.

Tennessee's ELVIS Act was the first US state law to shield a person's voice from AI replicas, adding voice cloning to the state right of publicity.

Some 46 to 48 US states now regulate deepfakes, with 28 targeting political deepfakes and 45+ covering NCII, forming a patchwork compliance environment.

A December 2025 executive order directs the DOJ and FTC to challenge state AI laws under preemption and Commerce Clause grounds, raising federal-state tension.

The FCC ruling places AI-generated and cloned voices under the TCPA, requiring prior express consent for AI voice calls to US residential and mobile lines.

The NO FAKES Act would create a federal right over a person's voice and visual likeness, with notice-and-takedown duties and a postmortem right up to 70 years.

The DEFIANCE Act would give victims of sexually explicit deepfakes a federal civil claim with liquidated damages up to $250,000 per violation.

The first US federal law on AI content harms criminalizes non-consensual intimate deepfakes and forces covered platforms to remove them within 48 hours.

Denmark's proposed Copyright Act amendment would treat a person's face, body, and voice as protected property, requiring consent for public digital replicas.

France's Loi SREN amended the Criminal Code to prohibit non-consensual deepfakes, with harsher penalties for sexual deepfakes under Article 226-8-1.

The Crime and Policing Bill 2025 would criminalize supplying nudification tools in the UK, targeting the deepfake supply chain rather than end users alone.

Section 138 makes creating or requesting non-consensual intimate images a UK criminal offence, whether or not the content is ever shared.

The Online Safety Act 2023 puts a duty of care on UK platforms to prevent intimate deepfakes, with Ofcom fines up to 10% of global revenue.

Directive 2024/1385 criminalizes non-consensual intimate deepfakes across the EU, with a member-state transposition deadline of 14 June 2027.

The DSA sets platform notice-and-action duties and annual systemic risk assessments for manipulated content and synthetic disinformation across the EU.

Article 50 of the EU AI Act requires machine-readable marking of AI output and visible deepfake labels, with transparency duties enforced from 2 August 2026.