A group of top experts and Nobel laureates has made a loud call for stricter rules around artificial intelligence. They want governments to set clear boundaries on what AI can and can’t do. The idea is to prevent AI from causing serious or unacceptable risks worldwide.

In a speech at the United Nations General Assembly, Nobel Peace Prize winner Maria Ressa urged nations to agree on “red lines” for AI. She emphasized the need for international rules that are enforceable and can be put into action by the end of 2026. More than 200 well-known figures, including industry leaders, Nobel winners, and former heads of state, signed the open letter backing this initiative.

What AI Activities Should Be Banned?

The campaign suggests banning AI in some very dangerous areas. For example, it calls for restrictions on AI used in nuclear command systems or autonomous weapons capable of killing without human oversight. Mass surveillance using AI that invades privacy is also on the list. They want to stop AI that tricks people by pretending to be human without revealing its true nature, which could mislead or deceive users.

The group also wants to prevent cyberattacks launched by AI, such as releasing malicious software that could disrupt vital infrastructure. Another concern is AI systems that can replicate or improve themselves without human control, which could spiral out of control if left unchecked. They stress that any international treaty should be clear about what’s forbidden, include verification mechanisms to ensure compliance, and establish an independent body to oversee enforcement.

Challenges in Making Global AI Rules Happen

Many experts worry about whether such rules can actually work. They ask if enough countries will support these bans and if they can be enforced effectively across borders. The concern is that the rules mainly target big AI vendors and hyperscalers—large companies offering powerful AI tools—rather than everyday users or smaller firms.

For example, these restrictions could impact how companies screen job applicants, decide on loans, or train AI models using sensitive customer data. But if a country signs on to the agreement, companies operating there would need to follow its rules, regardless of local laws. Countries like Germany, Canada, Switzerland, and Japan already have their own AI regulations, which might make new international rules less relevant.

Valence Howden, an AI adviser, says he supports the goal but questions how practical it is. “Risks aren’t tied to country borders,” he points out. He notes that the US is hesitant to regulate AI heavily, while even China talks about responsible AI use. Howden fears that the push to create global rules by the end of 2024 is too quick. He says that AI industry governance is moving slowly and warns we are nearing a point where AI becomes too hard to control.

Howden also doubts that big AI companies will follow any rules if they are adopted. “Can we trust large vendors to do this? No, they don’t now,” he says. Meanwhile, Brian Levine, a former federal prosecutor and expert in international standards, expects some agreement but doubts it will lead to real change. “Countries will agree in principle, but those principles will be vague,” he explains. Many nations will think, ‘It’s not really enforceable, so why worry?’ Past efforts, like trying to ban autonomous killer robots, haven’t achieved much.

Peter Salib, a law professor, adds that today’s AI systems are much more tangible and threatening than the robotic weapons discussed years ago. Still, he’s skeptical about the new push. “Most countries don’t care enough or want to give up sovereignty,” he says. Without strong enforcement, these rules risk being just words, not actual protections.

In the end, creating effective, global AI regulations remains a huge challenge. While many agree that some form of oversight is necessary, turning that into enforceable law is another story. The coming months will show whether nations can come together to set meaningful boundaries for AI’s future.

Inspired by

• https://www.computerworld.com/article/4061898/the-united-nations-attempt-to-regulate-ai-could-complicate-enterprise-compliance.html

Sources

• gmpg.org
• nobelprize.org
• red-lines.ai
• red-lines.ai
• red-lines.ai