The AI World Order, and India's place in it

Artificial Intelligence (AI) is on track to become the defining technology of the 21st century, according to the latest Stanford HAI AI Index, but its gains will be uneven without deliberate oversight. The data, it cautions, reveals that progress is accelerating faster than institutions can keep pace.

Technically, leading AI models are converging, with open-weight systems narrowing the gap. But the means of measuring that progress are fraying: benchmarks are saturating, transparency is declining, and independent tests do not always align with developer claims, the report notes. In science, AI is shifting from assisting discrete tasks to replacing entire workflows. In medicine, tools such as ambient scribes are scaling across health systems.

Regulation, meanwhile, is sending mixed signals. The European Union is enforcing its AI Act, the US is leaning towards deregulation, while Japan, South Korea and Italy are advancing national frameworks. Many developing economies, including India, are entering the arena, with AI sovereignty emerging as a unifying theme. Optimism is rising, but so is unease.

Here are some highlights from the study:

➡️ AI isn’t plateauing. Instead, it’s accelerating.

➡️ Over 90% of frontier models now come from industry, many matching or beating humans on advanced tasks. Coding benchmarks neared human parity in a year, while adoption hit 88%, with most students using generative AI.

➡️ The US-China AI gap has narrowed to near parity. China leads in AI publications, citations, and patents, while the US dominates in high-impact patents and output (50 notable models in 2025 vs China’s 30). South Korea tops patents per capita, and China’s share of the 100 most-cited AI papers rose from 33 in 2021 to 41 in 2024. In 2024, China accounted for 17.8% of AI publications in 2024, compared to 11.1% from Europe and 7.6% from India.

➡️ The US leads in AI data centres (5,427)—10 times more than any other country—but the chip supply chain is fragile. Most advanced chips rely on Taiwan’s TSMC, creating a critical single-point dependency despite US expansion.

➡️ AI shows gold-medal math performance, yet they are unable to read time from traditional clocks (with about 50% accuracy). AI Agents improved to about 66% task success but still failed one in three tasks.

➡️ AI models can outperform scientists in some domains, but scaling isn’t linear. They beat chemists on average, yet perform poorly in areas like astrophysics.

➡️ Robots thrive in labs but struggle at home—only 12% success rate. Simulations hit about 89%, highlighting the gap between controlled environments and real-world complexity.

➡️ The US leads AI investment ($285.9 billion) and startups, but talent inflows are falling sharply—down 89% since 2017, including an 80% drop in the last one year.

➡️ India leads in AI skill penetration at 3.0 (nearly triple the global average), followed by the US (2.0) and Germany (1.8), LinkedIn data shows. But gender gaps persist. In India, men list AI skills far more than women (3.1 vs 1.9), with a similar, slightly narrower gap in the US (2.1 vs 1.4).

➡️ AI’s environmental cost is rising fast. AI data centre capacity reached 29.6 GW, on a par with New York’s peak demand, while GPT-4o’s annual water use may exceed the drinking needs of 12 million people.

➡️ Smaller, specialised models often outperform much larger ones. Scientific AI is also more collaborative, unlike industry-led general AI.

➡️ Open-source AI continues to scale, with 5.6 million projects on GitHub and Hugging Face uploads tripling since 2023. India remains a growing contributor, representing 5.2% of all projects.

➡️ AI is reshaping clinical care, cutting documentation time and burnout, but strong evidence is limited. Nearly half of studies rely on exam-style data, and only about 5% use real patient data.

➡️ Education is lagging, even as AI use surges. More than 80% students use AI, but policies are unclear and uneven. Meanwhile, AI skills are growing globally, and more PhDs are staying in academia.

➡️ AI sovereignty is rising, with nations investing in domestic capability. Still, model development is concentrated in the US and China. Open source is widening participation and enabling more diverse models.

➡️ Regional efforts are building language-first AI from scratch, not waiting on global labs. Projects like SEA-LION and AI4Bharat are creating local data pipelines, tokenisers, and benchmarks. These initiatives make linguistic inclusion a core design goal, expanding responsible AI beyond major hubs.

➡️ India saw the sharpest rise in AI nervousness of any country surveyed.

➡️ Responsible AI is lagging capability. Safety reporting remains patchy: incidents rose to 362 (from 233 in 2024), and improving one dimension (for example, safety) can worsen another, like accuracy.

➡️ Experts and the public diverge sharply on AI’s impact—73% vs 23% positive views on jobs. Trust in AI governance is fragmented, with the US among the least confident in its own regulation.

AI TOOL OF THE WEEK

By AI&Beyond, with Jaspreet Bindra and Anuj Magazine

𝚃𝚑𝚎 𝙰𝙸 𝚏𝚎𝚊𝚝𝚞𝚛𝚎/𝚝𝚘𝚘𝚕 𝚠𝚎 𝚙𝚛𝚎𝚜𝚎𝚗𝚝 𝚝𝚘𝚍𝚊𝚢: 𝙲𝚕𝚊𝚞𝚍𝚎 𝚏𝚘𝚛 𝚆𝚘𝚛𝚍

𝚆𝚑𝚊𝚝 𝚙𝚛𝚘𝚋𝚕𝚎𝚖 𝚍𝚘𝚎𝚜 𝚒𝚝 𝚜𝚘𝚕𝚟𝚎? 𝙰 𝚕𝚊𝚠𝚢𝚎𝚛 𝚛𝚎𝚌𝚎𝚒𝚟𝚎𝚜 𝚊 𝚑𝚎𝚊𝚟𝚒𝚕𝚢 𝚛𝚎𝚍𝚕𝚒𝚗𝚎𝚍 𝚌𝚘𝚗𝚝𝚛𝚊𝚌𝚝 𝚊𝚝 5:00 𝚙𝚖, 𝚊𝚗𝚍 𝚝𝚑𝚎 𝚙𝚊𝚛𝚝𝚗𝚎𝚛 𝚠𝚊𝚗𝚝𝚜 𝚊 𝚌𝚕𝚎𝚊𝚛 𝚜𝚞𝚖𝚖𝚊𝚛𝚢 𝚘𝚏 𝚌𝚑𝚊𝚗𝚐𝚎𝚜, 𝚔𝚎𝚢 𝚛𝚒𝚜𝚔𝚜 𝚊𝚗𝚍 𝚙𝚛𝚘𝚙𝚘𝚜𝚎𝚍 𝚌𝚘𝚞𝚗𝚝𝚎𝚛-𝚎𝚍𝚒𝚝𝚜 𝚋𝚢 9:00 𝚊𝚖.

𝚃𝚑𝚒𝚜 𝚒𝚜 𝚗𝚘𝚝 𝚊 𝚏𝚛𝚒𝚗𝚐𝚎 𝚞𝚜𝚎 𝚌𝚊𝚜𝚎. 𝙳𝚘𝚌𝚞𝚖𝚎𝚗𝚝-𝚑𝚎𝚊𝚟𝚢 𝚙𝚛𝚘𝚏𝚎𝚜𝚜𝚒𝚘𝚗𝚊𝚕𝚜 𝚜𝚙𝚎𝚗𝚍 𝟹𝟶-𝟺𝟶% 𝚘𝚏 𝚝𝚑𝚎𝚒𝚛 𝚠𝚘𝚛𝚔𝚒𝚗𝚐 𝚍𝚊𝚢 𝚒𝚗𝚜𝚒𝚍𝚎 𝚆𝚘𝚛𝚍. 𝙲𝚕𝚊𝚞𝚍𝚎 𝚏𝚘𝚛 𝚆𝚘𝚛𝚍 𝚎𝚖𝚋𝚎𝚍𝚜 𝙰𝚗𝚝𝚑𝚛𝚘𝚙𝚒𝚌’𝚜 𝚊𝚜𝚜𝚒𝚜𝚝𝚊𝚗𝚝 𝚍𝚒𝚛𝚎𝚌𝚝𝚕𝚢 𝚒𝚗𝚝𝚘 𝙼𝚒𝚌𝚛𝚘𝚜𝚘𝚏𝚝 𝚆𝚘𝚛𝚍, 𝚛𝚎𝚊𝚍𝚒𝚗𝚐 𝚢𝚘𝚞𝚛 𝚏𝚞𝚕𝚕 𝚍𝚘𝚌𝚞𝚖𝚎𝚗𝚝, 𝚊𝚗𝚜𝚠𝚎𝚛𝚒𝚗𝚐 𝚚𝚞𝚎𝚜𝚝𝚒𝚘𝚗𝚜 𝚊𝚋𝚘𝚞𝚝 𝚜𝚙𝚎𝚌𝚒𝚏𝚒𝚌 𝚌𝚕𝚊𝚞𝚜𝚎𝚜 𝚠𝚒𝚝𝚑 𝚌𝚕𝚒𝚌𝚔𝚊𝚋𝚕𝚎 𝚌𝚒𝚝𝚊𝚝𝚒𝚘𝚗𝚜, 𝚎𝚍𝚒𝚝𝚒𝚗𝚐 𝚜𝚎𝚕𝚎𝚌𝚝𝚎𝚍 𝚝𝚎𝚡𝚝 𝚠𝚑𝚒𝚕𝚎 𝚙𝚛𝚎𝚜𝚎𝚛𝚟𝚒𝚗𝚐 𝚢𝚘𝚞𝚛 𝚏𝚘𝚛𝚖𝚊𝚝𝚝𝚒𝚗𝚐, 𝚊𝚗𝚍 𝚊𝚙𝚙𝚕𝚢𝚒𝚗𝚐 𝚎𝚟𝚎𝚛𝚢 𝚌𝚑𝚊𝚗𝚐𝚎 𝚊𝚜 𝚊 𝚝𝚛𝚊𝚌𝚔𝚎𝚍 𝚛𝚎𝚟𝚒𝚜𝚒𝚘𝚗 𝚢𝚘𝚞 𝚊𝚌𝚌𝚎𝚙𝚝 𝚘𝚛 𝚛𝚎𝚓𝚎𝚌𝚝 𝚒𝚗 𝚆𝚘𝚛𝚍’𝚜 𝚘𝚠𝚗 𝚛𝚎𝚟𝚒𝚎𝚠 𝚙𝚊𝚗𝚎.

𝙷𝚘𝚠 𝚝𝚘 𝚊𝚌𝚌𝚎𝚜𝚜: 𝚂𝚎𝚊𝚛𝚌𝚑 “𝙲𝚕𝚊𝚞𝚍𝚎 𝚋𝚢 𝙰𝚗𝚝𝚑𝚛𝚘𝚙𝚒𝚌” 𝚒𝚗 𝙼𝚒𝚌𝚛𝚘𝚜𝚘𝚏𝚝 𝙰𝚙𝚙𝚂𝚘𝚞𝚛𝚌𝚎. 𝙰𝚟𝚊𝚒𝚕𝚊𝚋𝚕𝚎 𝚘𝚗 𝙿𝚛𝚘, 𝙼𝚊𝚡, 𝚃𝚎𝚊𝚖, 𝚊𝚗𝚍 𝙴𝚗𝚝𝚎𝚛𝚙𝚛𝚒𝚜𝚎 𝚙𝚕𝚊𝚗𝚜．

𝙲𝚕𝚊𝚞𝚍𝚎 𝚏𝚘𝚛 𝚆𝚘𝚛𝚍 𝚌𝚊𝚗 𝚑𝚎𝚕𝚙 𝚢𝚘𝚞:

• 𝚁𝚎𝚟𝚒𝚎𝚠 𝚌𝚘𝚗𝚝𝚛𝚊𝚌𝚝𝚜 𝚏𝚊𝚜𝚝𝚎𝚛: 𝙰𝚜𝚔 𝚊𝚗𝚢 𝚚𝚞𝚎𝚜𝚝𝚒𝚘𝚗 𝚊𝚋𝚘𝚞𝚝 𝚢𝚘𝚞𝚛 𝚍𝚘𝚌𝚞𝚖𝚎𝚗𝚝 𝚊𝚗𝚍 𝚐𝚎𝚝 𝚊𝚗 𝚊𝚗𝚜𝚠𝚎𝚛 𝚕𝚒𝚗𝚔𝚎𝚍 𝚝𝚘 𝚝𝚑𝚎 𝚎𝚡𝚊𝚌𝚝 𝚌𝚕𝚊𝚞𝚜𝚎—𝚗𝚘 𝚖𝚘𝚛𝚎 𝚑𝚞𝚗𝚝𝚒𝚗𝚐 𝚝𝚑𝚛𝚘𝚞𝚐𝚑 𝚖𝚞𝚕𝚝𝚒𝚙𝚕𝚎 𝚙𝚊𝚐𝚎𝚜.
• 𝙴𝚍𝚒𝚝 𝚠𝚒𝚝𝚑 𝚌𝚘𝚗𝚏𝚒𝚍𝚎𝚗𝚌𝚎: 𝙴𝚟𝚎𝚛𝚢 𝙲𝚕𝚊𝚞𝚍𝚎 𝚜𝚞𝚐𝚐𝚎𝚜𝚝𝚒𝚘𝚗 𝚕𝚊𝚗𝚍𝚜 𝚊𝚜 𝚊 𝚝𝚛𝚊𝚌𝚔𝚎𝚍 𝚌𝚑𝚊𝚗𝚐𝚎 𝚢𝚘𝚞 𝚊𝚌𝚌𝚎𝚙𝚝 𝚘𝚛 𝚛𝚎𝚓𝚎𝚌𝚝—𝚏𝚞𝚕𝚕 𝚌𝚘𝚗𝚝𝚛𝚘𝚕, 𝚣𝚎𝚛𝚘 𝚜𝚞𝚛𝚙𝚛𝚒𝚜𝚎𝚜.
• 𝙲𝚛𝚘𝚜𝚜-𝚊𝚙𝚙 𝚌𝚘𝚗𝚝𝚎𝚡𝚝: 𝙺𝚎𝚎𝚙 𝚊𝚗 𝙴𝚡𝚌𝚎𝚕 𝚏𝚒𝚗𝚊𝚗𝚌𝚒𝚊𝚕 𝚖𝚘𝚍𝚎𝚕 𝚘𝚙𝚎𝚗 𝚊𝚕𝚘𝚗𝚐 𝚜𝚒𝚍𝚎 𝚢𝚘𝚞𝚛 𝚆𝚘𝚛𝚍 𝚖𝚎𝚖𝚘. 𝙲𝚕𝚊𝚞𝚍𝚎 𝚙𝚞𝚕𝚕𝚜 𝚍𝚊𝚝𝚊 𝚊𝚌𝚛𝚘𝚜𝚜 𝚋𝚘𝚝𝚑 𝚠𝚒𝚝𝚑𝚘𝚞𝚝 𝚌𝚘𝚙𝚢-𝚙𝚊𝚜𝚝𝚒𝚗𝚐.

𝙴𝚡𝚊𝚖𝚙𝚕𝚎:

𝙰 𝚕𝚊𝚠𝚢𝚎𝚛 𝚊𝚝 𝚊 𝚖𝚒𝚍-𝚜𝚒𝚣𝚎 𝚏𝚒𝚛𝚖 𝚛𝚎𝚌𝚎𝚒𝚟𝚎𝚜 𝚊 𝟹𝟾-𝚙𝚊𝚐𝚎 𝙽𝙳𝙰 𝚠𝚒𝚝𝚑 𝟷𝟸 𝚌𝚘𝚞𝚗𝚝𝚎𝚛-𝚙𝚊𝚛𝚝𝚢 𝚛𝚎𝚍𝚕𝚒𝚗𝚎𝚜, 𝚕𝚊𝚝𝚎 𝚘𝚗 𝚊 𝚆𝚎𝚍𝚗𝚎𝚜𝚍𝚊𝚢 𝚎𝚟𝚎𝚗𝚒𝚗𝚐. 𝚂𝚑𝚎 𝚘𝚙𝚎𝚗𝚜 𝚝𝚑𝚎 𝚍𝚘𝚌𝚞𝚖𝚎𝚗𝚝 𝚒𝚗 𝙼𝚒𝚌𝚛𝚘𝚜𝚘𝚏𝚝 𝚆𝚘𝚛𝚍, a𝚌𝚝𝚒𝚟𝚊𝚝𝚎𝚜 𝙲𝚕𝚊𝚞𝚍𝚎 𝚏𝚘𝚛 𝚆𝚘𝚛𝚍 𝚏𝚛𝚘𝚖 𝚝𝚑𝚎 𝚛𝚒𝚋𝚋𝚘𝚗, 𝚊𝚗𝚍 𝚝𝚢𝚙𝚎𝚜： “𝚂𝚞𝚖𝚖𝚊𝚛𝚒𝚜𝚎 𝚠𝚑𝚊𝚝 𝚝𝚑𝚎 𝚌𝚘𝚞𝚗𝚝𝚎𝚛-𝚙𝚊𝚛𝚝𝚢 𝚌𝚑𝚊𝚗𝚐𝚎𝚍 𝚊𝚗𝚍 𝚏𝚕𝚊𝚐 𝚊𝚗𝚢𝚝𝚑𝚒𝚗𝚐 𝚠𝚘𝚛𝚝𝚑 𝚍𝚒𝚜𝚌𝚞𝚜𝚜𝚒𝚗𝚐.”

𝙲𝚕𝚊𝚞𝚍𝚎 𝚛𝚎𝚝𝚞𝚛𝚗𝚜 𝚊 𝚐𝚛𝚘𝚞𝚙𝚎𝚍 𝚜𝚞𝚖𝚖𝚊𝚛𝚢 𝚚𝚞𝚒𝚌𝚔𝚕𝚢—𝚕𝚒𝚊𝚋𝚒𝚕𝚒𝚝𝚢 𝚌𝚑𝚊𝚗𝚐𝚎𝚜, 𝙸𝙿 𝚘𝚠𝚗𝚎𝚛𝚜𝚑𝚒𝚙 𝚎𝚍𝚒𝚝𝚜, 𝚝𝚎𝚛𝚖𝚒𝚗𝚊𝚝𝚒𝚘𝚗 𝚝𝚠𝚎𝚊𝚔𝚜 𝚠𝚒𝚝𝚑 𝚖𝚘𝚜𝚝 𝚊𝚐𝚐𝚛𝚎𝚜𝚜𝚒𝚟𝚎 𝚙𝚛𝚘𝚟𝚒𝚜𝚒𝚘𝚗𝚜 𝚛𝚊𝚗𝚔𝚎𝚍 𝚋𝚢 𝚜𝚎𝚟𝚎𝚛𝚒𝚝𝚢. 𝙴𝚊𝚌𝚑 𝚏𝚒𝚗𝚍𝚒𝚗𝚐 𝚒𝚜 𝚊 𝚌𝚕𝚒𝚌𝚔𝚊𝚋𝚕𝚎 𝚕𝚒𝚗𝚔 𝚝𝚑𝚊𝚝 𝚓𝚞𝚖𝚙𝚜 𝚍𝚒𝚛𝚎𝚌𝚝𝚕𝚢 𝚝𝚘 𝚝𝚑𝚎 𝚛𝚎𝚕𝚎𝚟𝚊𝚗𝚝 𝚌𝚕𝚊𝚞𝚜𝚎.

𝚂𝚑𝚎 𝚌𝚕𝚒𝚌𝚔𝚜 𝚘𝚗 𝚝𝚑𝚎 𝙸𝙿 𝚏𝚕𝚊𝚐. 𝚃𝚑𝚎 𝚌𝚘𝚞𝚗𝚝𝚎𝚛𝚙𝚊𝚛𝚝𝚢 𝚑𝚊𝚜 𝚚𝚞𝚒𝚎𝚝𝚕𝚢 𝚒𝚗𝚜𝚎𝚛𝚝𝚎𝚍 𝚕𝚊𝚗𝚐𝚞𝚊𝚐𝚎 𝚐𝚒𝚟𝚒𝚗𝚐 𝚝𝚑𝚎𝚖𝚜𝚎𝚕𝚟𝚎𝚜 𝚙𝚎𝚛𝚙𝚎𝚝𝚞𝚊𝚕 𝚛𝚒𝚐𝚑𝚝𝚜 𝚝𝚘 𝚍𝚎𝚛𝚒𝚟𝚊𝚝𝚒𝚟𝚎 𝚠𝚘𝚛𝚔𝚜.

𝚂𝚑𝚎 𝚝𝚢𝚙𝚎𝚜: “𝚁𝚎𝚠𝚛𝚒𝚝𝚎 𝚝𝚑𝚒𝚜 𝚌𝚕𝚊𝚞𝚜𝚎 𝚝𝚘 𝚛𝚎𝚜𝚝𝚘𝚛𝚎 𝚘𝚞𝚛 𝚜𝚝𝚊𝚗𝚍𝚊𝚛𝚍 𝚙𝚘𝚜𝚒𝚝𝚒𝚘𝚗 — 𝙸𝙿 𝚌𝚛𝚎𝚊𝚝𝚎𝚍 𝚍𝚞𝚛𝚒𝚗𝚐 𝚝𝚑𝚎 𝚎𝚗𝚐𝚊𝚐𝚎𝚖𝚎𝚗𝚝 𝚜𝚝𝚊𝚢𝚜 𝚠𝚒𝚝𝚑 𝚞𝚜, 𝚠𝚒𝚝𝚑 𝚊 𝚕𝚒𝚌𝚎𝚗𝚌𝚎 𝚝𝚘 𝚝𝚑𝚎 𝚌𝚕𝚒𝚎𝚗𝚝. 𝙰𝚙𝚙𝚕𝚢 𝚊𝚜 𝚊 𝚝𝚛𝚊𝚌𝚔𝚎𝚍 𝚌𝚑𝚊𝚗𝚐𝚎.”

𝙲𝚕𝚊𝚞𝚍𝚎 𝚖𝚊𝚔𝚎𝚜 𝚝𝚑𝚎 𝚎𝚍𝚒𝚝. 𝙸𝚝 𝚕𝚊𝚗𝚍𝚜 𝚒𝚗 𝚆𝚘𝚛𝚍’𝚜 𝚛𝚎𝚟𝚒𝚎𝚠 𝚙𝚊𝚗𝚎, 𝚎𝚡𝚊𝚌𝚝𝚕𝚢 𝚕𝚒𝚔𝚎 𝚊 𝚌𝚘𝚕𝚕𝚎𝚊𝚐𝚞𝚎’𝚜 𝚖𝚊𝚛𝚔𝚞𝚙. 𝚂𝚑𝚎 𝚛𝚎𝚟𝚒𝚎𝚠𝚜, 𝚊𝚌𝚌𝚎𝚙𝚝𝚜, 𝚊𝚗𝚍 𝚖𝚘𝚟𝚎𝚜 𝚘𝚗.

𝚆𝚑𝚊𝚝 𝚖𝚊𝚔𝚎𝚜 𝙲𝚕𝚊𝚞𝚍𝚎 𝚏𝚘𝚛 𝚆𝚘𝚛𝚍 𝚜𝚙𝚎𝚌𝚒𝚊𝚕?

• 𝚆𝚘𝚛𝚔𝚜 𝚒𝚗𝚜𝚒𝚍𝚎 𝚆𝚘𝚛𝚍, 𝚗𝚘𝚝 𝚋𝚎𝚜𝚒𝚍𝚎 𝚒𝚝: 𝙴𝚟𝚎𝚛𝚢 𝚎𝚍𝚒𝚝 𝚕𝚊𝚗𝚍𝚜 𝚊𝚜 𝚊 𝚝𝚛𝚊𝚌𝚔𝚎𝚍 𝚌𝚑𝚊𝚗𝚐𝚎 𝚒𝚗 𝚆𝚘𝚛𝚍’𝚜 𝚗𝚊𝚝𝚒𝚟𝚎 𝚛𝚎𝚟𝚒𝚎𝚠 𝚙𝚊𝚗𝚎, 𝚗𝚘 𝚗𝚎𝚠 𝚒𝚗𝚝𝚎𝚛𝚏𝚊𝚌𝚎 𝚝𝚘 𝚕𝚎𝚊𝚛𝚗.
• 𝙲𝚒𝚝𝚎𝚜 𝚒𝚝𝚜 𝚊𝚗𝚜𝚠𝚎𝚛𝚜: 𝙴𝚟𝚎𝚛𝚢 𝚛𝚎𝚜𝚙𝚘𝚗𝚜𝚎 𝚕𝚒𝚗𝚔𝚜 𝚍𝚒𝚛𝚎𝚌𝚝𝚕𝚢 𝚝𝚘 𝚝𝚑𝚎 𝚌𝚕𝚊𝚞𝚜𝚎 𝚒𝚝’𝚜 𝚛𝚎𝚏𝚎𝚛𝚎𝚗𝚌𝚒𝚗𝚐- 𝚗𝚘 𝚜𝚎𝚌𝚘𝚗𝚍-𝚐𝚞𝚎𝚜𝚜𝚒𝚗𝚐.
• 𝚁𝚎𝚊𝚍𝚜 𝚌𝚘𝚖𝚖𝚎𝚗𝚝 𝚝𝚑𝚛𝚎𝚊𝚍𝚜: 𝙲𝚕𝚊𝚞𝚍𝚎 𝚊𝚍𝚍𝚛𝚎𝚜𝚜𝚎𝚜 𝚛𝚎𝚟𝚒𝚎𝚠𝚎𝚛 𝚌𝚘𝚖𝚖𝚎𝚗𝚝𝚜, 𝚎𝚍𝚒𝚝𝚜 𝚝𝚑𝚎 𝚊𝚗𝚌𝚑𝚘𝚛𝚎𝚍 𝚝𝚎𝚡𝚝, 𝚊𝚗𝚍 𝚛𝚎𝚙𝚕𝚒𝚎𝚜—𝚝𝚞𝚛𝚗𝚒𝚗𝚐 𝚊 𝚖𝚞𝚕𝚝𝚒-𝚍𝚊𝚢 𝚛𝚎𝚟𝚒𝚎𝚠 𝚕𝚘𝚘𝚙 𝚒𝚗𝚝𝚘 𝚊 𝚜𝚒𝚗𝚐𝚕𝚎 𝚜𝚎𝚜𝚜𝚒𝚘𝚗.

𝙽𝚘𝚝𝚎: 𝚃𝚑𝚎 𝚝𝚘𝚘𝚕𝚜 𝚊𝚗𝚍 𝚊𝚗𝚊𝚕𝚢𝚜𝚒𝚜 𝚏𝚎𝚊𝚝𝚞𝚛𝚎𝚍 𝚒𝚗 𝚝𝚑𝚒𝚜 𝚜𝚎𝚌𝚝𝚒𝚘𝚗 𝚍𝚎𝚖𝚘𝚗𝚜𝚝𝚛𝚊𝚝𝚎𝚍 𝚌𝚕𝚎𝚊𝚛 𝚟𝚊𝚕𝚞𝚎 𝚋𝚊𝚜𝚎𝚍 𝚘𝚗 𝚘𝚞𝚛 𝚒𝚗𝚝𝚎𝚛𝚗𝚊𝚕 𝚝𝚎𝚜𝚝𝚒𝚗𝚐. 𝙾𝚞𝚛 𝚛𝚎𝚌𝚘𝚖𝚖𝚎𝚗𝚍𝚊𝚝𝚒𝚘𝚗𝚜 𝚊𝚛𝚎 𝚎𝚗𝚝𝚒𝚛𝚎𝚕𝚢 𝚒𝚗𝚍𝚎𝚙𝚎𝚗𝚍𝚎𝚗𝚝 𝚊𝚗𝚍 𝚗𝚘𝚝 𝚒𝚗𝚏𝚕𝚞𝚎𝚗𝚌𝚎𝚍 𝚋𝚢 𝚝𝚑𝚎 𝚝𝚘𝚘𝚕 𝚌𝚛𝚎𝚊𝚝𝚘𝚛𝚜.

AI BITS & BYTES

Now Hiring | Philosophers at Big Tech firms

Big Tech appears keen to signal a conscience—hiring philosophers to make the case, even as AI threatens to render such roles obsolete. This week, Google DeepMind brought on Henry Shevlin as a “philosopher”. Currently at Cambridge’s Leverhulme Centre for the Future of Intelligence, Shevlin works across cognitive science, AI ethics, and consciousness, with papers in journals like Journal of Consciousness Studies. His mandate now is to help machines think more responsibly.

At Anthropic, Amanda Askell is already doing just that—embedding ethical reasoning into systems like Claude. With a PhD in philosophy from NYU, Askell’s work spans AI alignment, decision theory, and formal epistemology, following earlier stints shaping safety research at OpenAI.

The irony is hard to miss. Even as tech firms recruit philosophers to teach AI right from wrong, some of their own leaders suggest those same systems could hollow out the humanities. Alex Karp, chief executive of Palantir Technologies, has consistently warned that AI will disrupt white-collar careers, especially in fields like philosophy, literature, law, and academia. Ironically, Karp himself has philosophy degrees from Haverford, Stanford, and a PhD from Frankfurt.

$1.35 million

It’s the average ransomware payout by organisations in India, which has emerged as the epicentre of ransomware activity in the Asia-Pacific, per Check Point. The latest report notes that 65% of affected Indian firms paid hackers during 2025.

Globally, attacks on manufacturers rose 56% from 937 incidents in 2024 to 1,466 in 2025 as hackers increasingly prioritise operational disruption and supply chain leverage over standalone data theft.

Meta may unseat Google as world’s largest ad player

Meta Platforms is expected to surpass Alphabet’s Google to become the world’s leading digital-advertising business—a first for the social-media company. Emarketer projects Meta’s ad revenue at $243.46 billion versus Google’s $239.54 billion.

The research firm’s estimates account for revenue after deducting traffic and other content acquisition costs, such as the money Google shares with its creators. Meta’s ad business is seeing a lift, thanks to new ad offerings, including the short-form video format Reels and the broader boost that AI has provided. Read more.

Is AI blocking creation of entry-level job?

A study by venture firm SignalFire found that hiring of candidates with less than one year of experience fell nearly 50% between 2019 and 2024, as work traditionally assigned to junior employees is being absorbed by AI systems.

IBM, in contrast, is increasing its hiring of Gen Z. “The companies three to five years from now that are going to be the most successful are those companies that doubled down on entry-level hiring in this environment,” Nickle LaMoreaux, IBM’s chief human resources officer, said during an interview with Charter this February.

🔗 AI, Iran war hit campus hiring in India

Meta’s AI reset is a threat for jobs, data and its future

Meta Platforms may have finally shown its hand on AI, but the bigger story isn’t the technology. Its new model, ‘Muse Spark’, is being positioned as the backbone of the company’s next phase, powering its AI app and soon embedding across Instagram, WhatsApp and Facebook.

Unlike previous headline-grabbing bets with the metaverse or smart glasses, this is a subtle shift. But it is also far more invasive. The company says it will use public posts to provide “context from your people, right where you need it”, eventually integrating Instagram Reels, photos and posts directly into responses. In effect, Meta is turning its social graph into training data and distribution at the same time. Read more.

🔗 ALSO READS

◾️ OpenAI’s leaked memo slams Anthropic’s approach to AI

◾️ The Algorithm of Cry and rise of babytech startups in India

◾️ How Oracle’s huge AI buildout is causing a cash crunch, layoffs

◾️ Anthropic Mythos AI should evoke concerns on sovereignty

◾️ Will smartphones survive the age of AI?

Tech Talk is a weekly newsletter by Leslie D'Monte on everything happening in the world of technology and AI. Want this delivered straight into your inbox? Subscribe here.