December 10th.IndiaProposed introduction of mandatory royalties, requirement AI CompaniesPaying for copyrighted content used in their model training, which may or will be reshaped OpenAI andGoogleCompanies operate in one of the most important and growing markets in the world。
THIS TUESDAY, LOCAL TIME, THE MINISTRY OF INDUSTRY AND INTERNAL TRADE PROMOTION (DPIIT) OF INDIA ISSUED A PROPOSED FRAMEWORK:AI WILL BE ALLOWED TO USE ALL COPYRIGHT-PROTECTED WORKS FOR MODEL TRAINING, SUBJECT TO THE PAYMENT OF ROYALTIES TO A NEW COLLECTIVE MANAGEMENT BODY COMPOSED OF RIGHTS-HOLDERS ORGANIZATIONS, WHICH WILL THEN BE ALLOCATED TO CREATORSI DON'T KNOW. THE PROPOSAL ARGUES THAT SUCH A “COMPULSORY LICENSING PACKAGE” MECHANISM COULD ENSURE THAT WRITERS, MUSICIANS, ARTISTS AND OTHER RIGHTS HOLDERS RECEIVE A REASONABLE REMUNERATION FOR THEIR WORK WHEN IT IS USED FOR TRAINING IN BUSINESS MODELS, WHILE REDUCING THE COSTS OF COMPLIANCE BY AI ENTERPRISES。
AT THE TIME OF THE INDIAN PROPOSAL, THERE WAS GROWING CONCERN IN THE GLOBAL MARKET ABOUT THE USE OF COPYRIGHT-PROTECTED MATERIAL TRAINING MODELS BY AI COMPANIES. IN THE UNITED STATES AND EUROPE, SUCH PRACTICES HAVE LED TO NUMEROUS LAWSUITS AGAINST RIGHTS HOLDERS SUCH AS AUTHORS, NEWS AGENCIES, ARTISTS AND OTHERS. COURTS AND REGULATORS ARE STILL ASSESSING WHETHER SUCH TRAINING CONSTITUTES “REASONABLE USE”, LEADING AI TO RAPIDLY EXPAND ITS OPERATIONS IN THE FACE OF LEGAL UNCERTAINTY AND TO A LACK OF A CLEAR REGULATORY FRAMEWORK。
IN CONTRAST TO THE ONGOING POLICY DEBATE BETWEEN THE UNITED STATES AND THE EUROPEAN UNION ON THE OBLIGATION OF TRANSPARENCY AND THE RATIONAL USE OF BORDERS, INDIA'S PROPOSAL IS ONE OF THE MOST INTERVENTIONIST INITIATIVES TO DATE, NAMELY, THE AUTOMATIC GRANTING OF RIGHTS TO AI COMPANIES TO USE COPYRIGHTED CONTENT SUBJECT TO COMPULSORY PAYMENT。
In late April, the Government of India established an eight-member committee to develop the relevant mechanisms. In the Commission ' s view, such a system would avoid years of legal controversy while ensuring that creators were paid from the outset。
IN A 125-PAGE OFFICIAL SUBMISSION, THE COMMITTEE DEFENDED THE SYSTEM AS THE LEAST BURDENSOME WAY TO MANAGE LARGE-SCALE AI TRAINING DATA, “TO PROVIDE EASY ACCESS TO CONTENT FOR AI DEVELOPERS ... TO REDUCE TRANSACTION COSTS ... AND TO ENSURE FAIR REMUNERATION FOR RIGHTS HOLDERS”. IT WAS ALSO NOTED THAT A SINGLE COLLECTIVE AUTHORITY WOULD SERVE AS A “ONE-STOP SHOP”, FREE OF THE CUMBERSOME PROCESS OF NEGOTIATIONS ON A CASE-BY-CASE BASIS AND ENSURE THAT BOTH REGISTERED AND UNREGISTERED CREATORS WOULD BE GRANTED ROYALTIES。
At the same time, the Committee stressed the importance of India as a market for generated AI (GenAI) tools. The paper cites the statement of OpenAI CEO Sam Altman that India is the second largest market for the company after the United States, “probably becoming our largest market”. On this basis, the Committee notes that, in view of AI ' s considerable income from Indian users and its reliance on the work of Indian creators to train their models, it would be reasonable to return some of its value to those creators. This is one of the core logics behind the establishment of a “balanced framework” to guarantee remuneration for creators。
THE INDIAN PROPOSAL IS AT A TIME OF GROWING LEGAL DISPUTES AROUND WHETHER AI CAN LEGALLY USE COPYRIGHT-PROTECTED CONTENT TRAINING MODELS。
In India, News Agency ANI has sued OpenAI in the High Court of Delhi for unauthorized use of the ANI articles. This case prompted the Court to examine whether AI training itself constituted “reproduction” or fell within the scope of “reasonable use”. Courts in the United States and Europe are facing similar disputes, with creators generally accusing the company of building its AI model based on unauthorized content。
However, not all agree with the model proposed by the Government of India。
The industry association Nasscom, representing technology enterprises such as Google and Microsoft, submitted a formal objection that India should adopt the broader Text and Data Mining exception to allow AI developers to use copyright-protected content for training subject to legal access. Nasscom warns:mandatory licensing systems may discourage innovation and suggest that dissenting right-holders should be allowed to opt out (opt-out) rather than impose corporate fees for all training data。
The Business Software Alliance (BSA), representing global science and technology enterprises such as Adobe, Amazonic Cloud Services (AWS) and Microsoft, also urged the Government of India to avoid the use of pure licensing. BSA called on India to introduce a clear TDM exception, stating that “Only relying on direct or statutory permission to regulate AI training data may be impractical and may not produce the best results”。
BSA FURTHER WARNS THAT LIMITING AI MODEL TRAINING TO SMALLER AUTHORIZED OR PUBLIC DOMAIN DATA SETS MAY LEAD TO A DECLINE IN MODEL PERFORMANCE AND “ADDITIONAL OUTPUT RESULTS REFLECT ONLY THE RISK OF TRENDS AND BIAS IN LIMITED TRAINING DATA”. IN CONTRAST, A CLEAR TDM EXCEPTION WOULD BETTER BALANCE TECHNOLOGICAL INNOVATION WITH RIGHTS-HOLDERS。
IN RESPONSE, THE COMMISSION OF THE GOVERNMENT OF INDIA INDICATED THAT IT HAD NOT ADOPTED THE “WIDESPREAD TDM EXCEPTION + OPT-OUT” MODEL ON THE GROUNDS THAT SUCH MECHANISMS EITHER WEAKENED COPYRIGHT PROTECTION OR WERE DIFFICULT TO IMPLEMENT EFFECTIVELY. INSTEAD, THE COMMISSION PROPOSED A “MIXED MODEL” WHEREBY AI COMPANIES CAN AUTOMATICALLY ACCESS ALL LEGALLY PUBLISHED COPYRIGHT-PROTECTED WORKS, BUT MUST PAY ROYALTIES TO THE CENTRAL COLLECTIVE ADMINISTRATION, WHICH DISTRIBUTES THE PROCEEDS TO CREATORS。
At present, the Government of India has initiated a public consultation process on the proposal, giving business and interested parties 30 days to submit their comments. The Commission will consolidate the feedback into final recommendations for submission to the Government for consideration。
As at 1AI, neither OpenAI nor Google had responded to requests for comments。