Raja Mohan: As Xi and Trump Open an AI Dialogue, India Faces Three Hard Questions

The news

New Delhi and Washington. Chinese President Xi Jinping meets U.S. President Donald Trump in Washington on Thursday, September 24, to begin a formal bilateral dialogue on artificial intelligence (AI) and its risks, C. Raja Mohan writes in his ‘Raja Mandala’ column in the Indian Express. The meeting comes barely 10 days after Mr. Xi used the Delhi BRICS summit to offer a Chinese vision of AI cooperation: he invited BRICS members to join the World Artificial Intelligence Cooperation Organisation (WAICO), launched by China in Shanghai this summer, proposed an open-source AI community for the forum, and presented China as the champion of developing countries seeking affordable, open-source AI, whose designs anyone may use and adapt. In Washington, Raja Mohan argues, the talk will be about the other side of AI: the dangers of powerful ‘frontier’ models, meaning the most advanced systems, and the need to prevent accidents, miscalculation and misuse. It is, he notes, their third summit in under 10 months. He traces the dialogue to Joe Biden and Mr. Xi’s meeting at Woodside, California, in November 2023; a year later at Lima the two affirmed that humans, not AI, must control decisions on the use of nuclear weapons. Under Mr. Trump, AI diplomacy has become more transactional, entangled with chips, export controls, rare earths and tariffs; when the two met in Beijing in May they discussed AI ‘guardrails’ but no arrangement followed. After talks in New York last weekend between Treasury Secretary Scott Bessent and Chinese Vice-Premier He Lifeng, Washington said the sides considered a standing AI dialogue and a notification system for incidents with national-security consequences. Raja Mohan poses three questions for India: how to navigate the contest while being part of the U.S.-led Pax Silica initiative on technology and supply chains, staying out of WAICO yet sitting in a China-dominated BRICS; how to compete when China can offer the Global South a wider technology package; and whether U.S.-China ‘strategic stability’ could become discriminatory rules for everyone else, as the nuclear order did. His prescription: trusted partners abroad, indigenous capability at home. The same day, a PTI report says Beijing called Taiwan a “red line” ahead of Mr. Xi’s three-day visit, the leaders’ second meeting since May, and is expected to oppose a proposed $11 billion U.S. arms package for Taiwan that Mr. Trump has yet to approve. At the UNGA, ahead of a Security Council meeting on AI and international security, Mr. Trump said he would not “stifle” a technology “bigger than Industrial Revolution”. Syllabus: GS2, effect of great-power policies on India; GS3, science and technology.

Static syllabus linkage

1. The nuclear order is the precedent that makes India wary of great-power technology rules. The Nuclear Non-Proliferation Treaty (NPT) was opened for signature in 1968 and entered into force in 1970; it defines a nuclear-weapon state as one that manufactured and exploded a nuclear device before January 1, 1967, which froze the club at five. India refused to sign, calling the treaty discriminatory. The Nuclear Suppliers Group (NSG) was formed in 1974 after India’s first nuclear test at Pokhran to restrict nuclear exports, and only in September 2008 did it grant India a special waiver following the India-U.S. civil nuclear initiative. This is the history Raja Mohan invokes when he warns that a U.S.-China ‘strategic stability’ on AI could harden into controls against others.
2. India has since joined most export-control regimes, but not the NSG. India became a member of the Missile Technology Control Regime in 2016, the Wassenaar Arrangement on conventional arms and dual-use goods and technologies in December 2017, and the Australia Group on chemical and biological weapons in January 2018. These regimes are informal groupings of supplier states that coordinate national export controls; they are not treaties. NSG membership still requires consensus, and China has blocked India’s entry by insisting on criteria for states outside the NPT. Membership matters because advanced AI chips are increasingly treated as dual-use goods under export controls.
3. Rivals have agreed rules before, and those rules bound everyone else. After the Cuban missile crisis, the U.S. and the Soviet Union set up a direct communication link, the ‘hotline’, in 1963. The Partial Test Ban Treaty of 1963 banned nuclear tests in the atmosphere, outer space and under water but not underground; India signed it in 1963. The Outer Space Treaty of 1967 barred nuclear weapons in orbit, and the Anti-Ballistic Missile Treaty and SALT I followed in 1972. The incident-notification system now being discussed by Washington and Beijing belongs to the same family of crisis-management tools.
4. India’s own AI and chip architecture, and its Taiwan position. The Union Cabinet approved the IndiaAI Mission in March 2024 with an outlay of ₹10,371.92 crore over five years, for public compute capacity, indigenous foundation models, datasets and skills. The India Semiconductor Mission was set up in December 2021 under a ₹76,000 crore programme to attract chip and display manufacturing. On Taiwan, India follows a ‘one China’ practice and has no diplomatic relations with Taipei, but has maintained the India-Taipei Association since 1995 for trade and cultural ties. The U.S. Taiwan Relations Act of 1979 commits Washington to provide Taiwan with arms of a defensive character, which is the legal basis of the $11 billion package Beijing opposes.

Why UPSC loves this

1. GS2 and GS3 now meet on technology geopolitics. The GS2 line on the ‘effect of policies and politics of developed countries on India’s interests’ now includes export controls, chip supply chains and AI rules, while GS3 covers ‘awareness in the fields of IT, computers and robotics’. Recent papers have asked about applications of AI; the next step is questions on AI governance and India’s place in it, for which this column is a ready frame.
2. Prelims has asked about export-control regimes and treaties. India’s membership of the MTCR, the Wassenaar Arrangement and the Australia Group, and its non-membership of the NSG, have been tested, as have the provisions of the NPT and the test ban treaties. The IndiaAI Mission and the India Semiconductor Mission, with their parent ministries and outlays, are likely scheme questions.
3. Interviews probe strategic autonomy in practice. Boards often ask whether India can be in BRICS and the Quad at once. Raja Mohan’s three questions give a candidate a precise answer: India is inside the U.S.-led Pax Silica, outside China’s WAICO, and still has to manage China-shaped BRICS rhetoric.

Analysis

1. Mr. Xi’s road from Delhi to Washington exposes the limits of BRICS’s AI rhetoric. In Delhi, Mr. Xi spoke of openness, access and diffusion; in Washington he will negotiate the rules of the frontier with the only other power that has one. Raja Mohan’s point is that multipolar language does not change the fact that technological power is bipolar. For India, which hosted the BRICS summit, this is uncomfortable: a forum it chaired became a platform for China’s AI offer to the Global South. The counter-view is that BRICS still gives India a seat where China’s proposals can be diluted, which it could not do from outside.
2. An incident-notification channel is the right first step, and India needs one of its own. In a crisis, as Raja Mohan notes, the first need is not a grand bargain but the ability to establish what happened, whether it was deliberate, and who is responsible. That is the logic of the 1963 hotline, and it applies to AI-enabled systems that may one day misread a signal on a disputed border. India faces China along the Line of Actual Control and Pakistan across the Line of Control, yet no AI-specific notification arrangement exists with either. The objection is that confidence-building with China has eroded since the 2020 clashes in eastern Ladakh; but that is precisely why low-cost technical channels are worth proposing.
3. The ‘condominium’ fear is real, but U.S.-China rivalry is also India’s leverage. The worry is that two powers that agree on strategic stability may also agree on who else gets advanced chips and models, as the NPT once divided the world into haves and have-nots. Raja Mohan himself offers the counterweight: competition for markets and for partners limits such collusion and creates room for emerging powers. The difference from the 1970s is that India is now inside a U.S.-led technology arrangement, Pax Silica, rather than outside every club. The risk flips: membership may bring its own compliance demands, such as aligning with American export controls, which India must negotiate rather than assume.
4. The Global South technology package is where India is losing ground. India’s emphasis on AI for developing societies resonated at the Delhi AI summit earlier this year, but China can bundle compute, infrastructure, hardware and advanced models. India’s strongest export is digital public infrastructure, such as Aadhaar-based identity and UPI payments, which many countries admire. Pairing that infrastructure with affordable AI services is India’s natural offer. The weakness is scale of compute and capital; without them, India offers blueprints while China offers finished systems.
5. Capability, not declarations, will decide India’s place, and no one is going to slow the race. Mr. Trump’s statement at the UNGA that he will not ‘stifle’ AI shows that the U.S. will not trade speed for safety, and Mehta notes on the same page that race incentives arrive exactly when the technology needs slowing. A global safety regime is therefore unlikely soon, and rules will be written by those with frontier capability. That makes the IndiaAI Mission and the India Semiconductor Mission strategic, not merely industrial, programmes. Raja Mohan’s prescription — trusted partners abroad, indigenous strength at home — is sound, but the gap he admits with America and China means the partnerships will carry most of the weight for years.

Administrator's brainstorm