The amendments are likely to stipulate that platforms using artificially intelligent algorithms or language models to train their machines must be devoid of any form of bias
The Ministry of Electronics and Information Technology is set to announce amendments prohibiting AI algorithms or language models with biases or training on biased datasets from open use
There will be a legal obligation on platforms to ensure all these parameters are followed
The government is reportedly likely to amend the Information Technology (IT) Rules of 2021 and introduce rules for regulating artificial intelligence (AI) companies and generative AI models.
The amendments are likely to mandate that platforms using artificially intelligent algorithms or language models to train their machines must be devoid of any form of ‘bias’, ET reported, citing sources familiar with the matter.
The Ministry of Electronics and Information Technology is expected to soon announce the amendments, which will mandate that AI algorithms or language models with inherent biases or trained on biased datasets be prohibited from launching for open use.
Apart from the safeguards for preventing bias in Al and language models, the IT Rules amendments may also bring new rules for deepfake, synthetic content and explicit instructions for platforms on loan apps, the report further said.
“The parameters for bias that we have right now are centred around caste, religion, community and national security, among other parameters. There will be a legal obligation on platforms to ensure all these parameters are followed, thereby expanding the ambit of trust and safety of the internet,” a senior government official told ET.
According to the official, the government will engage in consultations to identify and address additional parameters that contribute to bias, beyond those already recognised, to enhance safety measures.
Another official told ET that these AI models must undergo sandbox and stress testing to guarantee the absence of biases before introducing the tool to the market.
Earlier, Minister of State for Electronics and IT, Rajeev Chandrasekhar, said that the Digital India Act, intended to replace the 23-year-old IT Act of 2000, is not expected to be finalised before the upcoming general elections in 2024.
The Digital India Bill is designed to ensure a secure and accessible internet for Indian users while also seeking to diminish the dominance of major tech corporations. In contrast to the IT Act of 2000, which provides safe harbour provisions for intermediaries, the Digital India Bill places responsibility on intermediaries for the content published on their platforms.
In March 2023, MeitY released its presentation on the proposed Digital India Act, 2023, emphasising the “urgent need for a specialised and dedicated adjudicatory mechanism for online civil and criminal offences”.
As per earlier reports, the Centre was looking to release the first draft of the Digital India Bill by the end of July 2023. Among several new proposals, the government is expected to impose penalties of up to INR 500 Cr for any breaches of the provisions under the Bill.
Under the Bill, the Centre is also expected to provide clear definitions for offences such as doxing, cybersquatting, astroturfing, dog-whistling, and others.