India is preparing to explore a more comprehensive framework for artificial intelligence, with the government considering stronger safeguards against deepfakes, cybersecurity threats and harmful AI applications. The proposed approach could affect technology companies, startups, businesses and everyday users, although the final rules are still awaited.
India’s Proposed AI Law Puts Safety Under Focus
India’s artificial intelligence regulation plans have entered an important phase, with the Centre preparing a consultation paper on a possible new AI law. On October 28, 2026, NDTV reported that the government was examining safeguards for advanced AI models, including systems capable of identifying software vulnerabilities, enabling sophisticated cyberattacks or generating convincing synthetic content.
Union Electronics and Information Technology Minister Ashwini Vaishnaw had previously said that the government would release a consultation paper in the coming weeks. The proposed framework is expected to address AI safety, cybersecurity, deepfakes, potential harm to users and the need to keep humans in control of the technology.
The consultation paper is expected to invite views from citizens, technology companies, researchers and other stakeholders. Its publication would be an important step towards developing a broader regulatory approach, but it would not automatically mean that a new AI law had come into force.
The central challenge for policymakers is to protect people from harmful uses of artificial intelligence without restricting legitimate innovation or making AI development unnecessarily difficult for Indian businesses.
Why India Is Considering Stronger AI Safeguards
Artificial intelligence is being used in customer service, education, healthcare, financial services, content creation and software development. These applications can improve productivity, reduce routine workloads and help businesses serve customers more efficiently.
However, increasingly capable AI systems also introduce risks. Deepfakes can make fabricated audio and video appear authentic, while automated tools can be misused to create fraudulent messages, impersonate individuals or support cyberattacks.
The spread of synthetic content is particularly concerning when it involves financial scams, public misinformation or non-consensual images. A convincing video or voice recording can damage someone’s reputation or persuade a victim to transfer money before the material is verified.
The government is also examining risks from advanced AI models that can perform complex tasks with limited human supervision. Such systems may interact with software, process sensitive information or execute actions on behalf of users.
Existing laws and platform rules already address several forms of online harm. However, policymakers are considering whether additional provisions are needed to deal with risks created by the AI systems themselves, rather than focusing only on the content distributed through digital platforms.
Frontier AI Models Could Face Additional Safety Checks
One significant issue under consideration is how India should regulate frontier AI models. These are highly advanced systems at the leading edge of AI capabilities, which may present greater risks because of their ability to perform complex tasks or be misused.
According to NDTV’s October 28 report, officials are examining options such as pre-release assessments, testing in controlled environments and restrictions on particularly risky capabilities.
Under a possible pre-release assessment system, developers of certain advanced models could be expected to evaluate their systems for vulnerabilities before making them widely available. Testing in a sandbox, which is an isolated and controlled environment, can help researchers identify dangerous behaviour without exposing real-world systems to unnecessary risks.
Another question is how regulators should determine which models qualify for additional scrutiny. Training computing power could be one measure, but it may not capture every risk. Assessments based on actual capabilities may also be necessary.
These measures remain under consideration, and their final scope will depend on the government’s consultation and subsequent policy decisions.
For Indian developers, the distinction will be important. A framework that targets genuinely high-risk systems could improve safety without imposing identical requirements on every small AI application or research project.
Deepfakes and Online Fraud Could Face Tighter Controls
Deepfakes are likely to remain a major focus of India’s AI policy discussions. AI tools can generate realistic images, videos and voices, making it increasingly difficult for ordinary users to distinguish authentic material from fabricated content.
The potential harm extends beyond misleading social media posts. Criminals may use cloned voices to impersonate relatives, company executives or public officials. Fraudsters can also create fake investment promotions, fabricated endorsements and misleading videos to gain people’s trust.
A new framework could examine whether AI-generated content needs clearer identification, how platforms should handle complaints and what responsibilities developers and distributors should have when their systems are used to cause harm.
However, labelling alone may not solve the problem. A manipulated video can spread quickly before a warning is added, and genuine content can also be presented in a misleading context. Effective safeguards may therefore require a combination of identification tools, reporting mechanisms, enforcement and public awareness.
For users, basic precautions will remain important. People should verify unusual payment requests through a separate communication channel, avoid sharing sensitive information with unverified contacts and check important claims against reliable sources.
For businesses, employee training and stronger identity-verification procedures can reduce the risk of AI-assisted impersonation and fraud.
What New AI Rules Could Mean for Indian Businesses
Businesses are among the biggest potential beneficiaries of AI, but they may also face additional responsibilities if stronger safeguards are introduced.
Companies using AI for customer support, recruitment, financial decisions or internal operations may need to examine how these systems handle personal information, produce recommendations and respond to errors. The requirements would depend on the final rules and the nature of each application.
Organisations could also need clearer internal policies covering employee use of AI tools, data access, human review and incident reporting. For example, a company using an AI assistant to process customer records would need to consider whether confidential information is being shared appropriately and whether the output is checked before it influences a decision.
Startups may face a different set of challenges. Large technology companies generally have more resources to conduct safety testing and hire specialist teams. Smaller businesses and early-stage developers may find extensive compliance requirements more expensive.
This is why the definition of high-risk AI systems will matter. Proportionate rules could focus the strictest safeguards on applications that present significant risks while allowing lower-risk tools to continue developing under appropriate oversight.
Businesses should not assume that every possible requirement is already mandatory. They can, however, begin reviewing their AI use, documenting risks and establishing procedures for checking outputs and handling complaints.
Human Oversight and Accountability Will Be Important
Another question is who should be held responsible when an AI system causes harm. Potentially relevant parties include the developer that creates a model, the company that deploys it, the platform that provides access and the organisation that uses its output.
Responsibility may differ depending on the circumstances. A developer may need to address a known security weakness, while a business using AI to make customer decisions may need to ensure that its process includes appropriate checks.
Human oversight is especially important when AI systems can take actions rather than simply generate answers. An automated agent that accesses files, interacts with websites or executes software commands may create risks if it receives excessive permissions or operates without suitable monitoring.
Possible safeguards could include limiting access to sensitive systems, requiring human approval for high-impact actions, monitoring unusual behaviour and providing a way to stop an AI agent when necessary.
The objective should be to make responsibility clear without treating every error as the same kind of failure. The final framework will need to consider the capabilities of the system, the role of each party and the harm involved.
Can India Balance AI Innovation With Public Safety?
India’s approach to AI regulation will influence both domestic technology development and the way international AI companies operate in the country.
Strong safeguards can help build public trust, particularly in sectors such as healthcare, finance and education, where unreliable outputs can have serious consequences. Clear expectations can also help businesses understand their responsibilities before deploying AI at scale.
At the same time, overly broad restrictions could increase costs for startups, universities and independent researchers. If compliance becomes too expensive or uncertain, smaller developers may struggle to compete with established companies.
A practical framework would need clear definitions, proportionate obligations and transparent enforcement. It should also account for differences between low-risk productivity tools and systems capable of causing significant harm.
India’s existing legal framework, including the Information Technology Act, 2000, and applicable intermediary rules, already governs aspects of digital activity. The proposed AI framework would need to clarify how any new obligations interact with these existing provisions.
For businesses and users, the immediate priority is to follow confirmed legal requirements while preparing for possible changes. The consultation process should provide a clearer picture of the government’s direction once the paper is published.
Key Takeaways
- India is preparing a consultation process for a broader AI regulatory framework focused on safety, cybersecurity and harmful applications.
- Advanced AI models could face additional testing or safeguards, but the exact requirements have not been finalised.
- Deepfakes, financial fraud, human oversight and responsibility for AI-related harm are among the issues under consideration.
- Businesses can prepare by reviewing their AI systems, protecting sensitive data and introducing appropriate human checks without assuming that proposed measures are already law.
Frequently Asked Questions
1. Is India introducing a new AI law in October 2026?
The government has announced plans to issue a consultation paper on a possible new AI law. As of October 28, 2026, the final legislation and its requirements had not been confirmed in the reports reviewed.
2. Will the proposed AI rules affect small businesses and startups?
Potentially, yes. The impact will depend on the final framework, the type of AI application and the obligations assigned to developers and users. Risk-based rules could help prevent unnecessary burdens on smaller businesses.
3. Could AI-generated videos and deepfakes face stricter regulation?
Deepfakes are among the areas being considered. Possible measures include clearer identification of synthetic content, stronger complaint mechanisms and safeguards against harmful uses. The final requirements remain undecided.
4. What should Indian businesses do before the new rules arrive?
Businesses can review where AI is being used, protect confidential information, verify important AI-generated outputs, limit access to sensitive systems and establish procedures for reporting errors or security incidents. They should monitor official announcements for confirmed legal obligations.
Leave a comment