Give Data, Take AI
India needs cutting-edge AI. It should use data as bargaining chip for unrestricted access to US models
Trump admin’s order to block top-tier AI access to non-Americans is not surprising, but it is problematic on two fronts. One, it hurts global collaboration. Two, for middle powers like India, and many smaller countries, it exposes dangers of dependence. Let’s be clear, US leads world in AI by an enormous margin. China is second, in what is essentially a two-horse race. Analysts say its best models are “months behind” America’s. But in a race that really started less than four years ago, with ChatGPT’s public release, “months” is a long time.
For now, the US restriction applies to Anthropic’s Mythos 5 and Fable 5 models. Fable 5 was released just last week. It’s the public version of the highly capable Mythos, with stronger safeguards built in. But US govt appears to believe these might be bypassed. If that happens, Fable 5 could be used to cause grave harm. For example, it could discover vulnerabilities in a govt’s computer systems, and carry out cyberattacks. US fears its own systems, including defence systems, could come under attack. Hence, the abrupt ban.
But this sets a precedent. From now on, US, and also China, may reserve their best AI models for their own use, on national security grounds. The rest will wait for hand-me-downs. But AI is more than software. It’s a vital tool for manufacturing, healthcare, scientific research, etc. Without the latest and best models – accessed for a price, of course – other nations can’t compete. How can they overcome this disadvantage?
There are three ways. One, join hands to develop AI together. Since AI development is costly, resources – chips, data centres, energy, research – can be pooled. But wide cooperation of this nature is easier preached than practised. ‘Sovereign AI’ is the second way – develop your own AI ecosystem, from chips to foundational models and apps. Europe is getting serious about this. India has talked about it long enough, but cost is a hurdle. Where Alphabet, Amazon, Meta, and Microsoft intend to spend $650bn on AI this year, India got just 0.6% of global AI funding last year. It’s the reason why Taiwan and S Korea just overtook us in market capitalisation.
The third way is the most pragmatic. Remain in the US camp, use its AI, but with some leverage. What does India have that US AI giants want desperately? User data. India should make access to its data conditional – for example, with a data localisation mandate – so that such AI setbacks don’t happen.
https://www.anthropic.com/news/claude-fable-5-mythos-5
Disclaimer
Views expressed above are the author’s own.
