The Stalemate on AI Safety: Why Global Regulation Remains Elusive

• Geopolitical Rivalry Blocks Regulation: The United States and China view each other as primary competitors in AI, making a joint slowdown or pause in development politically impossible.
• Leadership Prioritizes Speed Over Safety: Both Donald Trump and Xi Jinping have publicly stated that slowing AI progress is unnecessary, with Trump explicitly aiming to maintain US dominance.
• Limited Legislative Action in the US: Congressional recesses and political divisions have stalled significant safety bills, despite warnings from prominent industry figures.
• Divergent Industry Stances: While some AI pioneers advocate for strict governance, influential tech leaders like Nvidia’s Jensen Huang dismiss doomsday scenarios as irresponsible fearmongering.
• European Influence Wanes: Despite its history of tech regulation, the EU has struggled to gain traction in global AI safety discussions compared to US and Chinese dynamics.

The narrative surrounding artificial intelligence has shifted dramatically from panic to pragmatic competition. Last week, global attention focused on the potential for an AI apocalypse, but this anxiety is being overshadowed by the strategic interests of the two nations with the most advanced AI industries: the United States and China.

Donald Trump and Xi Jinping are scheduled to meet in Washington DC next Wednesday. While their discussions will likely touch upon AI governance, early indicators suggest neither nation is willing to implement regulatory constraints that might hinder their progress. In recent days, both leaders have stressed that a slowdown in AI development is unnecessary. Trump has framed the issue as a matter of national supremacy, stating, “We’re leading China in AI... I want to keep it that way, because whoever wins AI, wins.”

This stance is reinforced by Silicon Valley leadership. Many tech executives are wary of Chinese ambitions and support aggressive US advancement. Even Dario Amodei, who has called for a global coordinated pause on certain AI developments, has characterized China as a significant threat, complicating any unified front against rapid development. Conversely, China interprets appeals for a slowdown as an attempt by the US to solidify its lead. From Beijing’s perspective, accepting such a proposal would be akin to handing an advantage to a leading opponent.

However, this does not mean China ignores safety concerns entirely. Last week, Beijing released the third iteration of its AI safety governance framework. This update addresses specific risks identified by global experts, including threats from autonomous agents, cybersecurity vulnerabilities in frontier models, and the dangers of self-improving AI systems. These are the same three frontiers highlighted by Amodei as critical areas for risk management.

The combative tone between the US and China does not preclude all diplomatic engagement. In the lead-up to the presidential meeting, US Treasury Secretary Scott Bessent met with his Chinese counterpart, He Lifeng. They described these talks as “very successful” and discussed the possibility of an international alert system for AI incidents, modeled after existing nuclear warning systems. The two delegations plan to meet again in two months to further this dialogue.

Bessent emphasized that the US proposal includes a notification mechanism between the two countries, aiming for a shared vision of common goals and threats. However, without mutual agreement on regulation, broader global standards remain unlikely. A decade of dire warnings has failed to slow the industry’s pace, and political responses appear to be following suit.

Trump has dismissed AI doomsday warnings as “a HOAX,” making it improbable that he would sign any safety bill presented to him. In Congress, legislative progress is equally stalled. Although Senator Bernie Sanders has promised to propose prohibitions on extremely capable AI, no such bill is expected to pass both chambers in the near future. House Speaker Mike Johnson recently dismissed the chamber for a seven-week midterm campaigning recess, effectively halting meaningful progress on AI safety legislation while the nation focuses on elections.

Representative Yassamin Ansari criticized this move on CNN, stating, “It is just beyond the pale that Mike Johnson has called off Congress... We’re going home for seven weeks. And during that time, we’re going to see this frontier AI technology continue to evolve, continue to cause damage.”

The AI industry itself is not united on the need for regulation. While Amodei and OpenAI’s Sam Altman have invited new restrictions and nearly reached a joint legally binding safety testing agreement, they represent only one faction within the business ecosystem.

Influential figures like Nvidia CEO Jensen Huang and adviser David Sacks hold significant sway with the Trump administration and advocate for accelerating AI development. Huang recently stated that there is a “0% chance” AI will destroy the world by 2030, calling such scare tactics unnecessary and irresponsible.

Europe, which has traditionally acted as a global tech regulator, has found its leverage diminished in this debate. Despite high concerns for safety, Ursula Von der Leyen’s invitations to frontier labs to discuss pacing development garnered little response from US or Chinese entities.

Historical precedents offer mixed signals. In 2023, Geoffrey Hinton, a Nobel laureate often called a “godfather of AI,” left Google due to fears that bad actors could use the technology to harm humanity. His resignation did inspire significant legislative action, including a US executive order and the EU’s AI Act.

He expressed optimism about public movement toward safety. This optimism was bolstered when the Canadian and German governments announced funding of up to £160m for his non-profit organization, which aims to create an “honest AI” to act as a guardrail against rogue agents. While such funding provides tangible support for safety research, the broader political landscape remains dominated by competitive acceleration rather than coordinated restraint.