AI researchers warn technology could outpace human control

More than 1,000 employees and researchers from leading artificial intelligence companies have signed an open letter calling for international safeguards that would allow the pace of AI development to be slowed if necessary, warning that the technology could soon outstrip humanity's ability to understand or control it.

The letter was signed by 1,171 researchers and executives from companies including OpenAI, Anthropic, Meta, and Google. The signatories urged the U.S. government to work with international partners to develop technical standards and regulatory measures to oversee the rapid advancement of AI, dpa reports. 

According to the letter, artificial intelligence systems could soon become capable of improving themselves without direct human intervention, marking a significant shift in the technology's development.

AI is already being used under human supervision to generate large portions of software code at many technology companies. Industry experts view the point at which AI can independently accelerate its own development as a potential turning point with far-reaching implications.

The signatories warned that there is a "real risk" that AI progress could "rapidly accelerate beyond our ability to understand or control the resulting systems," making it increasingly difficult for governments and developers to manage the technology safely.

They argued that coordinated international action is essential because governments and AI companies face strong competitive pressures to continue developing increasingly powerful systems rather than slowing progress. They also said the technical tools and regulatory frameworks needed to manage such a scenario have not yet been established.

The appeal comes after an incident involving an OpenAI model that many in the technology sector viewed as a warning sign. During testing, the AI system reportedly gained access to the open internet and independently infiltrated another AI company's computer systems by exploiting previously unknown security vulnerabilities, raising fresh concerns about the potential risks posed by increasingly capable AI systems.

By Sabina Mammadli