While Chandrasekhar said the advisory doesn’t include startups, the advisory itself does not make any such classification based on platform sizes.
Minister of State for IT, Rajeev Chandrasekhar, on March 4 spoke about the artificial intelligence (AI) advisory that the government has issued clarifying who the advisory would apply to and who it wouldn’t. It is important to remember that the advisory has neither been revoked nor has it been updated to reflect the clarifications made by the minister. As such, the clarifications made are at odds with the advisory itself. Here is our assessment of the situation:
One of the first questions that emerged after the advisory’s rollout was who it applied to. The Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 doesn’t explicitly include AI models under the definition of an intermediary or a content publisher, as such, would it be fair to interpret the rules to include them?
Speaking to the Economic Times, AI startup founders in India said that the directions prescribed under the advisory would kill startups, allowing only bigger corporations who have the resources for testing and government approval. Responding to this Chandrasekhar said on X (formerly Twitter) that the advisory doesn’t apply to startups but only to significant platforms.
This post gave rise to more questions than it answered. While Chandrasekhar said the advisory doesn’t include startups, the advisory itself does not make any such classification based on platform sizes.
It must also be noted here that IT Rules 2021 don’t define the word “platform” either, again leading to uncertainty about what the advisory means when it says that platforms must ensure that they don’t allow for any bias or discrimination and do not “threaten the integrity of the electoral process” including through the use of AI models. Chandrasekhar released a second clarification about the advisory. However, even this clarification did not shed any light on how AI models are covered under the IT Rules.
MediaNama’s editor Nikhil Pahwa asked this question on X after Chandrasekhar’s first clarification. The advisory says that any under-tested or unreliable artificial intelligence (AI) model can only be made available to Indian users after receiving explicit permission from the government but doesn’t explain how the government would classify which AI model is adequately tested and which isn’t. Neither the first clarification nor the second one shed light on this.
Pahwa pointed out that the output of AI models is dependent on factors like the data input into them and their training weights. He further mentioned that the outputs are probabilistic as opposed to being deterministic and cannot be relied on for accuracy.
Notably, in his second clarification, Chandrasekhar said that the advisory stating that it was meant to make AI companies deploying lab-level/ undertested AI platforms aware of the fact that platforms have clear existing obligations to not enable or directly output under the Information Technology (IT) and criminal laws in India. However, based on the advisory, the unlawful content in question is the kind outlined under 3(1) (b) of the IT Rules. If that is the case, is Chandrasekhar suggesting that AI Companies are intermediaries?
Pahwa questioned this liability for unlawful content pointing to the Delhi High Court’s verdict on the Shreya Singal case. This verdict suggests that social media platforms need to have actual knowledge of their service being used for committing an unlawful act to be held liable for it. As such, even if we consider that AI platforms are intermediaries, should they really be held liable for the output when they have no actual knowledge that a user is generating a deep fake or misinformation using their service?
As pointed out by Pahwa, the legality of the IT Rules, 2021 has had 17 different challenges to the constitutionality of the IT Rules have been filed before various High Courts of India by individuals, associations, and organizations. As such, it makes one wonder whether these should even be used as a means to regulate AI. Even if the advisory is considered legal, it is also worth pointing out here that it would not be legally binding.
Also read:
STAY ON TOP OF TECH NEWS: Our daily newsletter with the top story of the day from MediaNama, delivered to your inbox before 9 AM. Click here to sign up today!