Why India’s competition regulator is warning companies about AI – Crypto News – Crypto News
Connect with us
Why India’s competition regulator is warning companies about AI Why India’s competition regulator is warning companies about AI

Metaverse

Why India’s competition regulator is warning companies about AI – Crypto News

Published

3 weeks ago

on

India’s competition watchdog has stressed the need to prevent anti-competitive practices in the use and development of artificial intelligence (AI). In a report released on Monday, the Competition Commission of India (CCI) has proposed a self-audit for companies to keep track of how they use AI and the data they train AI models on. Implementing such a practice comes with multiple caveats. Mint explains:

How will the AI self-audit work?

The CCI’s indicative framework to companies suggests that any enterprise using AI should document the “decision-making process” of both the main algorithm and the sources of data. It specifies that companies building AI models should ensure that they have built-in safeguards against anti-competitive recommendations, verifiable via audits at regular intervals. The CCI also suggested that AI models should not share “commercially sensitive data” with a business’s competitors and any pricing strategy determined by AI would be reviewed if the regulator suspects discriminatory practices. The watchdog will host workshops, advocacy sessions and eventually set up a think tank comprising “academics, technologists and public policy experts” to monitor AI use in enterprises.

Does this target Big Tech in India?

Not exactly, at least for now. Three lawyers and policy consultants said that the AI regulatory framework may not apply to Big Tech companies. Big Tech “typically designs its compliance policies with the most stringent regime in the world, i.e. the European Union (EU), and is unlikely to make its massive AI models work differently in India”, said Rahul Rai, partner and cofounder of law firm Axiom5 Law Chambers. Experts also added that the market study is a precursor to a potential regulatory framework for companies in AI, but for now, it is unlikely to be included in future renditions of regulating Big Tech in India.

Does CCI have the talent to vet AI?

Questions have been raised about how CCI would impose AI regulations. A senior official privy to discussions said that the regulator’s approach is not proactive, where a set of rules would be prescribed. The CCI will likely work with external advisors and consultants with subject matter expertise, something it has proposed as part of its think tank. The think tank will offer advice on AI-related investigations depending on the cases being pursued or complaints received.

How will AI regulation work, and for whom?

Rai said while the CCI report is not a binding legal framework, companies may follow its recommendations as a good practice in general. In future, if a company is investigated, having followed a self-audit of AI will give it a basis to either rebut allegations of anti-competitive conduct or seek a reduction in potential penalties. The regulatory approach is to dissuade companies that may seek to train large AI models to scour a market for pricing data and significantly undercut competitors to gain market share, to eventually control an industry. This would imply to Indian companies building foundational models, as well as enterprises using AI for business strategy.

Can explainability in AI really work?

The history of Big Tech and artificial intelligence has suggested that explainable AI—which provides internal workings and reasonings that humans can understand—has never taken off commercially, even as the EU pursued a strong regulatory approach against AI algorithms, also deemed to be black boxes. If imposed as a regulation, explainability in well-established large language models such as those by Big Tech firms would be difficult to achieve. This could have complicated commercial implications, which is why the CCI’s approach will be to ask Indian enterprises using AI models based on their own data to bring explainability to the system. Still, AI needs massive troves of data to be trained and explainability is complex, tedious and expensive to attain.

Related Topics:
Continue Reading

Trending