AI outpaces rules as industry leaders and EU warn of risks

Calls for restraint from figures inside the tech industry have collided with a geopolitical rush to dominate artificial intelligence, leaving regulators and governments scrambling to keep up. In recent days prominent voices — including Dario Amodei of Anthropic and public supporters such as Sam Altman, Elon Musk and Demis Hassabis — urged greater caution as the most capable models advance rapidly.

The alarm is grounded in concrete developments. At the start of September OpenAI said its model "Astra" had crossed a threshold the company describes as a critical level of cyber capabilities. Anthropic's September report also described instances where AI was used in cyber operations, fraud, surveillance and other abuses, "including operations with a high degree of autonomy." The company added that, in those cases, "people still made the most important decisions, such as choosing the targets and using the results."

The central dilemma is strategic: slowing development may reduce technological risk, but unilateral restraint can hand competitive advantage to rivals. That tension is reinforced by official policy. The American Action Plan for Artificial Intelligence under the Trump administration frames the technology as a global race and prioritises speeding innovation, building infrastructure and strengthening U.S. leadership — including tighter controls on exports of advanced chips linked to new military capabilities.

Power, investment and the infrastructure behind the race

China treats AI as a national priority too. Its "AI plus" plan published in 2025 sets targets for deep integration into science, industry, consumer products, public services and governance, aiming for applications of the new generation of intelligent devices and agents to exceed 70 percent by 2027 and 90 percent by 2030. Behind the geopolitical contest sits a vast economic engine: the Stanford AI Index for 2026 reports private AI investment in the U.S. in 2025 reached about $285.9 billion, compared with $12.4 billion in China. Goldman Sachs has estimated cumulative global AI-related consumption could equal roughly 5 percent of world GDP by the end of the decade.

There are knock-on effects for energy, construction and markets. Morgan Stanley analysts say nearly $200 billion of data-centre projects in the U.S. were cancelled or delayed during the last year and the first quarter of 2026, in part because of domestic political and local resistance. That makes any idea of slowdown not only a technological or ethical question, but one that would ripple through investment, jobs and supply chains.

Meanwhile Europe has moved to bind innovation into law. From 2 August 2026 parts of the EU Artificial Intelligence Act entered into force, introducing obligations on transparency, labelling of AI-generated or altered content, and rules for general-purpose models. European Commission President Ursula von der Leyen warned members of the European Parliament that while AI can boost productivity and quality of life, it also brings serious risks if left unchecked.

"I am optimistic that artificial intelligence has the potential to help humanity if we use it in the right way. However, like any powerful new technology, alongside all the advantages, artificial intelligence also brings great risks. If we do not remove those risks, we will never be able to fully harness its potential," von der Leyen said, adding that "If those who develop this technology are so clear in their warnings, then we too must take them seriously." She also called for a Europe-level mechanism — likened to an "Article 4" — so that when one member state is threatened the whole bloc can respond.

Photo: press material from the event