AI Summit 2025: Safety concerns and an energy-guzzling future

Fifty-eight countries, including China, France and Australia, signed a joint statement on inclusive and sustainable artificial intelligence (AI) for people and the planet at the AI Action Summit in Paris on Tuesday, 11th February 2025.

This statement, which aims to make AI more accessible, human-centric, ethical, safe, secure and trustworthy, is the third international declaration on AI. It focuses on five key themes: public interest AI, the future of work, innovation and culture, trust in AI and global AI governance.

However, the statement was not signed by the United States or the United Kingdom.

To advance the human-centric vision for AI, countries such as France, India, and Germany have launched a public interest AI platform and incubator. This initiative seeks to reduce the divide between existing public and private AI initiatives and address the digital divide. It aims to co-create a trustworthy AI ecosystem by supporting technical assistance and capacity-building projects in areas such as data, model development, transparency, auditing, computing, talent financing and collaboration.

The third summit marked a shift from the safety concerns that dominated the previous two summits. It was stated that “excessive regulation” could hinder the development of emerging AI technologies.

US Vice President JD Vance further stated that, at this moment, we face the extraordinary prospect of a new industrial revolution, one comparable to the invention of the steam engine. However, he warned, this potential will never come to fruition if overregulation deters innovators from taking the necessary risks to drive progress.

In 2024, Meta founder and CEO Mark Zuckerberg, alongside 100 other experts, directly criticised Europe's slow progress in AI, writing an open letter that attributed the delay to excessive regulations.

Amidst the ambitious AI framework, journalist Shakeel Haseem noted in his newsletter *Transformer* that corporate executives dominated the summit, while the International Science of AI Safety report was relegated to a side event. The AI safety report, authored by 100 AI experts, is the world’s first comprehensive synthesis of current literature on the risks and capabilities of advanced AI systems.

AI is becoming an energy-intensive technology. A single GPT query, for example, demands around 2.9Wh of electricity—nearly ten times more than a Google search—and this is expected to escalate in the future. According to Goldman Sachs, power demand from data centres will grow by 160% by 2030. The social cost, in terms of carbon dioxide emissions from meeting data centre demands, is estimated to be between $125 billion and $140 billion.

The International Energy Agency (IEA) reports that data centres consumed 1.65 billion gigajoules of electricity in 2022, representing about 2% of global demand. By 2026, total electricity consumption from AI could exceed 1,000 terawatt-hours—roughly equivalent to Japan's total electricity consumption. The average power consumption of large data centres is around 100MW or more, which is the equivalent of the annual electricity demand of 350,000 to 400,000 electric cars.