Future of AI: Ultimate Choice of the Collective

In my last column, we looked at Artificial Intelligence and explored how it could potentially affect our electoral democracy. We don’t yet know what AI technology will mean for the future of society, but we should act now to build the tools that we need to ensure it serves us rather than oppresses us, and to protect people from its most dangerous applications. 

If we look for a framework within which to approach solutions to the problems posed by AI, we have the supporting strength of human rights laws from four general categories. These are: data protection rules to protect rights in the data-sets used to develop and feed artificial intelligence systems; special safeguards for government uses of artificial intelligence; safeguards for private sector uses of artificial intelligence systems; and investment in more research to continue to examine the future of artificial intelligence and its potential interferences with human rights. 

As for Indian elections, the Election Commission of India (ECI) issued an advisory to political parties and politicians as late as Monday, 6 May 2024 -- with two out of the seven phases already over, and the third to be held the next day -- and that too on account of the kerfuffle over the doctored video of Home Minister Amit Shah (see my previous column). Demanding responsible and ethical use of social media in election campaigning, the ECI asked political parties to remove fake content within three hours of its coming to the parties’ notice. 

The EC also warned against the misuse of AI-enabled tools that create deepfakes, distort information, and propagate misinformation, emphasising the need to uphold the integrity of the electoral process. It reminded political parties of the existing legal provisions that govern the regulatory framework against the use of misinformation and impersonation using deepfakes: the Information Technology Act, 2000 and the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules 2021, the Indian Penal Code, and the framework created by the Representation of People Act, 1950 and 1951, and the provisions of the Model Code of Conduct. 

Contrary to the perception that the AI space in India is so unregulated as to allow an unceasing deluge of defamatory and deceptive deepfakes, India’s IT rules already have a framework in place to contend with morphed images and videos. Needless to say, this framework needs to be tailored to the expanding powers and prevalence of AI and deepfake technology  such that their malicious use can be instantly checked. 

Under Rule 3(2) of the Information Technology Rules, 2021, intermediaries like website providers, social media platforms and video streaming services are required, within twenty-four hours from the receipt of a complaint (of impersonation, fakery, morphing etc), to remove or disable access to such content hosted, stored, published or transmitted by it. 

Of course, the aggrieved person can also approach the Cyber Crime Cell of their local police station, and/or report the content on the National Cybercrime Reporting Portal Online; complaints can also be lodged through certain provisions of the Information Technology Act, 2000. Furthermore, there are provisions in the IT rules (Sections 66E, 67, 67A, and 67B) for protecting the privacy of individuals and punishing miscreants for sharing obscene and sexually explicit material electronically. 

However, the devil lies in the implementation of the complaint mechanism available to ordinary citizens, which at times seems to exist merely on paper. Rather than being preoccupied with drafting new legislation, the Union Government first ought to focus on augmenting existing safeguards, ensuring that all deepfake complaints are acted on with the urgency that was on display in the Amit Shah doctored-video controversy. 

While the Union Ministry of Electronics and Information Technology (MEITY) came out with an advisory to nominate a “Rule 7” officer for assisting the aggrieved to file an FIR against the intermediaries and offender, there has been no movement in this regard. In March 2024, amid rising concerns of arbitrarily censoring AI tools, MEITY withdrew its mandate for AI models and LLMs to mandatorily seek the Union Government’s approval before rolling them out for Indian users. Now such programmes can be made available to Indian users only after “appropriately labelling the possible inherent fallibility or unreliability of the output generated.”

Despite having assured the Union Government that they will work closely with it and Indian voters to protect their access to truthful information, big tech giants like Google, Meta and Open AI remain complacent, their cavalier attitude allowing for deep-faked videos to remain online without cautionary labels of any kind. While YouTube promised to implement rules requiring creators to disclose the presence of artificial content in their creations, it has not yet delivered on this promise, even as three out of seven phases of the Indian general elections are over. Meta has also outlined plans to identify and label AI-generated content, as well as deploy a fact-checking team to combat misinformation -- but this will only begin post-elections! 

Policymakers from Brussels to Washington are striving to draft legislation restricting the use of AI-powered audio, images and videos on the campaign trail. Racing ahead of the United States, the EU has drafted a trailblazing Artificial Intelligence Act, aimed at regulating the use of AI in the EU; however, it will come into effect only after the EU’s June 2024 parliamentary election. European Commission Vice President Vera Jourova said she has sent letters to key political parties in European member states with a “plea” to resist using manipulative techniques. However, she said, politicians and political parties will face no consequences if they do not heed her request. 

In the United States Congress, bipartisan legislation was introduced in September 2023 to ban the use of materially deceptive AI-generated content in elections. "The Protect Elections from Deceptive AI Act" seeks to clamp down on falsely depicting federal candidates using AI; unfortunately, it is unlikely to become a law before the November 2024 Presidential election. For want of a federal law, sundry states in the United States have enacted laws penalizing those making deceptive videos about politicians, creating a patchwork of policy across the country. Such laws can guide India in its pursuit to tame the tempests of AI and deepfake technology.

Another important fundamental right that can be easily impacted by AI is freedom of speech and expression, given the growing dependence on AI systems for the control of content online. We are seeing this already in areas such as Sentiment Analysis, identification of fake news, and AI applications in everyday life, ranging from smart assistants to the wretched auto-correct technology on our mobile devices. Our right to free speech is greatly vulnerable to violation in systems such as automated content removal used by many social media organisations and online sites, considering that technology has a restricted ability to identify the context and tone of what we are saying. 

For its part, Open AI has thus far refrained from officially rolling out the Voice Engine, its text-to-voice model that merely needs prompting and a single audio byte of around fifteen seconds in order to generate natural sounding versions of human voices speaking across languages, with and without accents. “We recognise,” said Open AI in a statement, “that generating speech that resembles people’s voices has serious risks, which are especially top of mind in an election year.” Be that as it may, numerous other voice cloning tools are now a few clicks away, with one available on the Apple App Store itself. 

Artificial Intelligence, in other words, is more Artificial than Intelligent. And it is susceptible to manipulation by intelligent but malicious human beings, as we have seen in some political parties’ use of bot-enabled online accounts that act as real users and issue programmed responses to create particular choices and opinions for others. This is a real hazard to genuine free expression, and it is already being used around us.

This is why it is so necessary to establish suitable frameworks that can protect human rights in the current stages of the development of Artificial Intelligence. While it cannot be denied that the technology is still in its nascent stages we should work to find solutions to ensure that the emerging technologies are designed, developed, deployed and regulated in a way that enable, rather than threaten, human rights. 

Our aim should be to select the right, timely approach that neither impedes AI’s capacity to facilitate our economic growth and alleviate some of the world’s most unrelenting problems, nor allows AI to flourish at the expense of our human rights. The ultimate choice will emerge from the preferences of businesses, governments, and individuals like us -- it is we who will collectively determine how the AI ecosystem will evolve in future. Let us all think about it carefully today so that we get it right tomorrow.