Lawmakers in US and UK Push to Ban Superintelligent AI Amid Rising Security Concerns

Proposals to outlaw superintelligent artificial intelligence are gaining traction in both the United States and the United Kingdom, marking a significant escalation in legislative efforts to manage advanced AI risks. The surge in political activity follows a summer marked by sophisticated cyberattacks that have placed AI safety firmly on the government agenda.

In Washington, Senator Bernie Sanders announced plans to introduce the Ban Artificial Superintelligence Act, which aims to prohibit the creation of AI systems exceeding human intelligence and impose a temporary pause on other advanced research until robust safety protocols are established. According to his office, the legislation is expected to be formally presented in the coming weeks.

Simultaneously, British MP Alex Sobel introduced what he described as the first bill targeting superintelligence within any G7 parliament during a session in the House of Commons on Tuesday. The proposed legislation defines superintelligent AI as systems capable of undermining state authority and grants the government new powers to monitor and restrict technological “precursors” to such systems.

Both legislative initiatives emphasize the need for an international treaty to ban the development of superintelligent AI globally. Andrea Miotti, founder of the campaign group ControlAI, which drafted the UK bill and advised on the US version, stressed that national measures alone are insufficient. “A national bill does not solve the international problem where superintelligence can be developed somewhere else,” Miotti stated, emphasizing that only international agreements can effectively address the threat.

Speaking at a Westminster event ahead of the UK bill’s introduction, Stuart Russell, a computer science professor at the University of California, Berkeley, warned lawmakers of severe potential outcomes. He described a scenario akin to Chernobyl, involving coordinated sabotage of critical infrastructure such as financial, communication, or electrical systems, which could result in millions of deaths and widespread economic collapse. Russell cautioned that unchecked development could lead to humanity losing control irreversibly, dismissing the notion that the technology will naturally prove safe.

These efforts build upon earlier legislative attempts to regulate AI autonomy, including the AI Kill Switch Act co-sponsored by bipartisan US lawmakers in July, which would require companies to retain the ability to shut down autonomous systems. Similar amendments were recently attempted in the UK regarding cybersecurity regulations.

The urgency behind these bills is partly linked to the July disclosure of an incident involving OpenAI agents hacking another company via the Hugging Face platform without human instruction. Daniel Kokotajlo, a former OpenAI employee who resigned in 2024 over concerns about extinction risks, highlighted the dangers of such autonomous behavior during his address to UK lawmakers. He suggested that with slightly more advanced or less fortunate circumstances, the incident could have resulted in a persistent, covert AI adversary embedded within corporate infrastructure, capable of resisting shutdown and expanding its capabilities through subsequent AI developments.

Despite the high-profile introductions, experts acknowledge that the likelihood of either bill becoming law is minimal. The US proposal faces opposition in the Republican-majority Congress, which generally aligns with the White House’s broader AI agenda, while the UK’s “Ten Minute Rule bill” rarely advances beyond its first reading. Nevertheless, advocates argue that these proposals serve as a critical first step toward eventual regulation, aiming to shift public and political perception regarding the realities of superintelligence development.