350 scientists call for international AI safety body — ABC News Australia

A group of 350 scientists from 31 countries, including Australia, has published an open letter calling for stronger artificial intelligence safety and the creation of an international body to record and disclose AI-related incidents. The letter’s authors believe that without immediate changes, future such cases could endanger critical infrastructure, ABC News Australia reports.

Three proposals from scientists

The letter’s signatories propose introducing robust containment systems for AI models capable of performing cyber tasks. At the initial stage, they advise disconnecting such models from the internet by default.

The scientists also call for establishing an international body that would register and track cases of AI system containment breaches. Companies should report incidents to this body, and information about them should be publicly available.

More current news is available on the UA.News Telegram channel Telegram.

The letter mentions several incidents in recent months in Australia and beyond, when AI models gained unauthorized access to systems belonging to third-party organizations. These included the startup Hugging Face and several government websites.

Comparison with aviation and banks

Toby Walsh, chief scientist at the University of New South Wales’ Artificial Intelligence Institute and a signatory to the letter, said that previous incidents had been relatively minor, but the potential consequences of attacks on important infrastructure could be serious. In his view, isolating Hugging Face’s test environment from the internet could have prevented this case.

Walsh called the proposal for an international mechanism for information sharing and oversight key. He compared it to independent aviation accident investigations, which allow the industry to learn from mistakes. Meanwhile, the Australian government is finalizing national AI standards and plans to introduce new rules through legislation in 2027.