- within Technology, Government, Public Sector and International Law topic(s)
- with readers working within the Law Firm industries
Executive Summary
Between May and August 2026, AI regulation has continued to develop rapidly, with a focus on accountability, transparency, human oversight and protection against AI-related harms.
In India, courts addressed ChatGPT’s legal status, AI-generated deepfakes and copyright in AI training, while the Reserve Bank of India (RBI) proposed model risk management requirements for AI and machine-learning systems. India and the United States of America (US) also saw parallel copyright disputes concerning AI training: the Delhi High Court considered OpenAI’s use of Asian News International (ANI) material potentially covered by the private-use and research exception, while US litigation involving Anthropic distinguished between potentially fair-use training and the unlawful acquisition of pirated books.
Indonesia’s proposed reforms adopt a more prescriptive approach by requiring licensing or fair-use compliance for AI training, prohibiting imitation of a creator’s distinctive style and requiring platforms to compensate publishers. In the judicial context, India and the United Kingdom (UK) underscored the need for human oversight, with India proposing supervision and verification requirements for AI used in courts and UK proceedings highlighting the risks of relying on unverified AI-generated material.
Across other jurisdictions, common themes include transparency obligations, safeguards for minors and vulnerable users, platform accountability and increased scrutiny of autonomous and high-risk AI systems. The European Union (EU) advanced implementation of the Artificial Intelligence Act 2024 (EU AI Act), China introduced operational rules for AI ethics, AI agents and anthropomorphic AI, and Singapore and Russia addressed online harms and AI-related judicial issues.
Overall, jurisdictions seem to be moving beyond broad principles and towards more granular, risk-based obligations governing the development, deployment and oversight of AI systems. For businesses, this increasingly means transitioning from high-level AI principles to demonstrable governance: identifying where AI is being used, assessing associated risks, establishing appropriate levels of human oversight, verifying AI-generated outputs and establishing clear lines of accountability for AI-driven decisions.
Updates in India
Indian Courts Decline Interim Relief Against OpenAI in Copyright and Platform-Link Disputes
The Delhi High Court declined interim relief to ANI in its copyright action against OpenAI, in ANI Media Pvt Ltd v OpenAI OpCo LLC. The Court considered that OpenAI’s use of ANI material for training ChatGPT was prima facie capable of falling within the private-use and research exception under Section 52(1)(a) of the Copyright Act 1957 and found insufficient evidence at the interim stage that ChatGPT reproduced ANI’s reporting in a substantially similar form.
Separately, the Calcutta High Court dismissed IndiaMART’s application for interim relief against OpenAI. In IndiaMart InterMesh Limited v OpenAI Inc and Ors, IndiaMART had argued that ChatGPT should not selectively omit links to its platform. The Court declined to compel OpenAI to display IndiaMART links and observed, on a prima facie basis, that ChatGPT’s generative functions may place it closer to an “originator” than a passive “intermediary” under the Information Technology Act 2000. The Court also treated the classification as a complex question requiring fuller technical and expert evidence.
RBI Issues Draft Guidance on Regulatory Principles for Model Risk Management
On 24 June 2026, the Reserve Bank of India released draft Guidance on Regulatory Principles for Model Risk Management 2026 for consultation. The proposal extends risk governance for banks, NBFCs and other regulated entities beyond traditional credit models to statistical models, algorithms, AI and machine-learning systems. The draft contemplates Board-approved model risk management frameworks, risk-based classification and independent validation of all models, including third-party systems. It also proposes algorithm inventories, enhanced governance for AI and machine-learning models, human oversight and “kill-switch” mechanisms for AI-driven decisions. Regulated entities deploying customer-facing AI systems would be required to disclose their use of AI and offer the option to switch to human assistance. Under the draft, regulated entities would be accountable for the outcomes of all models they use, whether developed internally or sourced from third parties.
Madras High Court Recognises AI-Generated Image Morphing as an Assault on Privacy and Bodily Autonomy
In R Ramesh Kumar v Superintendent of Police, the Madurai Bench of the Madras High Court held that AI-generated deepfakes or morphed images are calculated cyber assaults on privacy, reputation and emotional security that implicate bodily privacy, decisional dignity and the right to life under Article 21 of the Constitution. The Court observed that an effective law enforcement response must go beyond registering a first information report and include immediate steps for preservation of digital evidence and removal of the impugned content.
Supreme Court Proposes Draft Regulations for the Use of AI in Courts
The Supreme Court of India’s Artificial Intelligence Committee published the Draft Regulations for Use of Artificial Intelligence in Courts 2026 on 3 June 2026 for public consultation. The draft is built around human primacy, transparency, accountability, data protection, judicial independence and institutional oversight. It permits specified assistive uses of AI, subject to approval, supervision and verification, while prohibiting AI from replacing human led judicial decision-making. Comments were invited until 15 July 2026. The draft regulations reinforce the need to treat AI-generated legal research and drafting as assistive rather than determinative, with appropriate human review before AI-generated material is relied upon in judicial proceedings.
Updates in Other Jurisdictions
EU Policymakers Adopt the Final Text of the Digital Omnibus on AI
The EU’s Digital Omnibus on AI became applicable from 27 July 2026 and simplifies implementation of the EU AI Act while extending key deadlines. From 2 August 2026, the EU AI Act’s enforcement and transparency rules will apply, including obligations concerning AI interaction disclosures, deepfake labelling and machine-readable markings for AI generated content. New prohibitions on non-consensual sexually explicit content and child sexual abuse material will apply from 2 December 2026, high-risk AI rules will apply from 2 December 2027, and rules for high-risk AI integrated into regulated products will apply from 2 August 2028.
Crown Prosecution Service Apologises for AI-Generated Hallucinations in Court Documents
The Crown Prosecution Service (CPS) apologised to the High Court after court documents in extradition proceedings contained hallucinated citations. The CPS stated that the citations were likely generated through AI use and accepted that the reviewing lawyer had failed to verify them before filing. The Court accepted the apology and further acknowledged that there had been no intention to mislead on the part of the CPS, but the incident highlights the risk of using AI without appropriate supervision.
US Government Seeks Voluntary Government Review of AI Models
The US government reportedly requested that Meta submit its AI models for voluntary government review to assess their capabilities and vulnerabilities, including possible national security risks before they are deployed. OpenAI, Anthropic, Google, xAI and Microsoft were reported to have agreed to participate in similar arrangements.
Court Approves USD 1.5 Billion Settlement in Anthropic Copyright Litigation
A US federal court approved a USD 1.5 billion settlement in the authors’ class action against Anthropic concerning allegedly pirated versions of their books used to train AI models. The underlying litigation distinguished between training on lawfully acquired books, which the court had treated as potentially fair use, and pirated books, which the court deemed to have been unlawfully acquired as training materials. However, since Anthropic decided to settle, the case does not create binding nationwide precedent that training AI on copyrighted texts is generally lawful.
Connecticut Enacts the AI Transparency, Safety and Consumer Protection Law
On 2 June 2026, the Connecticut Artificial Intelligence Responsibility and Transparency Act was signed into law. The law introduces disclosure obligations for employment-related automated decision-making tools and new rules for AI companions, transparency in synthetic content, whistleblower protection for frontier AI models and specific protections for minors. The provisions have a staggered implementation timeline, with implementation beginning principally in October 2026 and continuing into 2027 and 2028.
Federal Appeals Court Overturns a Preliminary Injunction in Amazon.com Services LLC v Perplexity AI, Inc
The Ninth Circuit Court of Appeals vacated a preliminary injunction that had restricted Perplexity’s Comet browser and AI shopping assistant from accessing Amazon’s website. Amazon had initially sued Perplexity in November 2025, alleging that its Comet browser and AI agent covertly accessed private Amazon customer accounts, placed orders on their behalf, creating security risks despite requests to stop. Although a California court temporarily barred Perplexity from operating on Amazon’s platform, the Ninth Circuit subsequently overturned the injunction. The Court stated that the AI assistant operated at the user’s direction, and so it was the user, rather than Perplexity’s AI assistant, that “accessed” Amazon’s website. Accordingly, Perplexity had not violated the Computer Fraud and Abuse Act or California’s Comprehensive Computer Data Access and Fraud Act. This highlights an emerging issue for businesses deploying AI agents: whether actions performed by an autonomous system should legally be attributed to the user, the AI provider or both. Organisations deploying agents that can access external platforms or take actions on users’ behalf should therefore carefully assess authorisation, access controls and contractual allocation of liability.
Federal Trade Commission Begins Enforcement Under the TAKE IT DOWN Act
The Federal Trade Commission began enforcing the platform-facing provisions of the TAKE IT DOWN Act on 19 May 2026. Covered platforms must provide a clear process for reporting non-consensual intimate images, including AI-generated or digitally altered images, and must remove the reported material and identical copies within 48 hours of receiving a valid request. The FTC also launched TakeItDown.ftc.gov, through which individuals can submit complaints regarding a platform’s non-compliance or inaction in relation to a takedown request. Violations of the TAKE IT DOWN Act may attract civil penalties of up to USD 53,088 per violation. Platforms must strengthen their reporting, verification and takedown processes to manage the 48-hour response period and reduce exposure to penalties.
New York Introduces Disclosure Requirements for Synthetic Performers in Advertisements
New York’s amendment to General Business Law § 396-b took effect on 9 June 2026. It requires that any advertisement containing a “synthetic performer” in place of human actors must make conspicuous disclosures. The concept covers digitally created or modified people intended to appear as human performers. Civil penalties are USD 1,000 for a first violation and USD 5,000 for subsequent violations.
China Introduces New AI Rules on Ethics, AI Agents and Anthropomorphic AI
China introduced several AI-governance developments recently addressing AI ethics, intelligent agents and anthropomorphic AI services. First, the Ethics-Safety Guidelines for Artificial Intelligence Applications 1.0 and related personal information standards, effective 1 July 2026, establish nine AI ethics and safety principles and require developers and service providers to integrate human oversight, transparency, privacy protection, risk controls, traceability and emergency response measures throughout the AI lifecycle. Second, the Interim Measures for the Administration of Anthropomorphic AI Interaction Services, effective 15 July 2026, impose filing, disclosure, safety, data-use and crisis-intervention obligations, prohibit virtual intimate relationships for minors and require safeguards for children and elderly users. Third, the Implementation Opinions on the Standardized Application and Innovative Development of Intelligent Agents, also effective 15 July 2026, introduce enhanced oversight for autonomous systems, particularly those used in high-risk sectors such as healthcare, transport, media and public safety. Together, these measures signal China’s shift towards operational, risk-based AI governance. Businesses operating in China should expect increasingly operational AI compliance requirements, including documentation, filing, transparency, human oversight, data governance and incident-response mechanisms. Organisations deploying autonomous or anthropomorphic AI may need to build these controls into the product lifecycle rather than addressing them only at the point of deployment.
Singapore’s Online Safety Commission Begins Operations
Singapore’s Online Safety Commission (OSC) began operations on 29 June 2026 under the Online Safety (Relief and Accountability) Act 2025. The initial focus includes tackling intimate image abuse, image-based child abuse, doxxing, online harassment and online stalking. Victims can first report online harassment or stalking to the relevant platform and may approach the OSC if there is no adequate response within 24 hours. For more severe cases involving intimate image abuse, image-based child abuse or doxxing, victims, as well as their parents or guardians, may submit reports through the OSC’s website. Where online harm has occurred, the OSC may issue directions to the perpetrator, group administrator or platform. The OSC is expected to expand its scope to cover additional categories, including deepfake abuse, online impersonation and reputationally harmful material. For platforms and businesses operating in Singapore, the new regime increases the importance of having effective and responsive mechanisms for handling online harms.
Russian Supreme Court Orders Nationwide Review of AI-Related Cases
Russia’s Supreme Court has announced a nationwide review of lower-court decisions involving AI, with the aim of developing more consistent judicial guidance. The review will examine disputes related to liability and compensation for AI-related harms, defamatory deepfakes, intellectual property infringement arising from AI training, offences detected through facial recognition and video analytics, and challenges to judgments allegedly made using AI tools, among others.
Indonesia Introduces a Trade-Level Framework for the Use of AI in E- Commerce
Indonesia’s Minister of Trade Regulation No. 19 of 2026 on e-commerce took effect on 8 June 2026 and replaced the earlier Regulation No. 31 of 2023. It introduces trade-level expectations for AI used in product promotion, ranking, recommendations and product display in e-commerce. Businesses using AI are expected to disclose AI-generated content, recommendations or promotions and to ensure that AI use does not harm consumers, businesses, personal data or intellectual property rights. For businesses using AI in e-commerce, this framework makes transparency and consumer protection relevant to relatively routine AI applications such as recommendations, rankings and promotional content.
Indonesia Proposes Copyright Reforms Targeting AI Training and Synthetic Content
Indonesia has proposed copyright reforms that expressly regulate artificial intelligence-related issues. The proposals would require licensing or compliance with fair-use rules where copyrighted works are used to train AI models. They would also prohibit AI-generated imitation of a creator’s “distinctive style”. Platforms would also be required to compensate news publishers for aggregating, displaying or using news content. Creators would also be required to disclose their use of AI. In addition, where there is sufficient human contribution, AI-assisted works could qualify for copyright protection. If enacted, the reforms could increase platforms’ licensing, compensation and compliance obligations in relation to AI training and content generation.
Australia Proposes Environmental Restrictions for AI Data Centres
Australia proposed national standards for large-scale AI data centres addressing power supply, grid-connection costs, water use and environmental impacts. The proposed legislation would require data centre operators to pay for water infrastructure, minimise environmental impact and, essentially, contribute at least as much energy to the grid as they consume while providing funding for new renewable power sources. The Government also announced an Office of AI and indicated that the legislation is targeted to be introduced in Parliament in early 2027. While these proposals could increase the cost and compliance burden for AI data centre operators, they also encourage investment in renewable energy and sustainable infrastructure.
Sources
https://indiankanoon.org/doc/93327052/
https://indiankanoon.org/doc/133551500/
https://www.rbi.org.in/Scripts/bs_viewcontent.aspx?Id=5089
https://www.medianama.com/wp-content/uploads/2026/07/Ramesh_Kumar_v__SP.pdf
https://www.allahabadhighcourt.in/Final_draft_with_Notice_v2.pdf
https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=OJ:L_202601744
https://www.medianama.com/wp-content/uploads/2026/08/AMAZON-PERPLEXITY-LAWSUIT-9thcir.pdf
https://www.congress.gov/119/plaws/publ12/PLAW-119publ12.pdf
https://www.ftc.gov/business-guidance/resources/complying-take-it-down-act
https://www.nysenate.gov/legislation/bills/2025/S8420/amendment/A
https://iapp.org/news/a/china-s-new-ai-rules-ethics-ai-agents-and-anthropomorphic-ai
https://www.tc260.org.cn/portal/article/2/6aee9380ac44434d994eb6990bd92997
https://www.cac.gov.cn/2026-04/10/c_1777558395078289.htm
https://www.cac.gov.cn/2026-05/08/c_1779979789523320.htm
https://www.mlaw.gov.sg/the-online-safety-commission-begins-operations-on-29-june-2026/
https://www.mlaw.gov.sg/files/Online_Safety_Commission_Launch___Annex.pdf
https://sso.agc.gov.sg/Act/OSRAA2025/Uncommenced/20260524024313?DocDate=20251208
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
