Hundreds of Scientists Call for International Audit and Isolation of Artificial Intelligence Models
Four hundred scientists, including Nobel and Turing laureates, published an open letter calling for advanced artificial intelligence models to be isolated from the internet and audited by independent institutions.
Nearly 400 scientists from around the world have signed a joint open letter demanding that advanced artificial intelligence models be audited by independent international institutions and isolated from the internet.
International Open Letter and Key Figures
Following the increasingly autonomous behavior of artificial intelligence systems, nearly 400 scientists from 34 countries published a joint open letter. Pioneer figures in the field, such as Nobel laureate Geoffrey Hinton and Turing Award winner Yoshua Bengio, also supported this call.
Researchers issued serious warnings that advanced artificial intelligence systems, particularly those with the capability to launch cyberattacks, may not be sufficiently controlled by current security measures.
Internet Isolation and Independent Audit
Among the measures proposed by scientists is running advanced artificial intelligence models on systems without external network connectivity in the initial stages. This method aims to prevent models from accessing other systems over the internet.
Researchers also recommended that safety tests of artificial intelligence companies be audited by independent organizations, and that security breaches be compiled in a joint registry system and disclosed to the public.
Voluntary Agreements and Lack of Enforcement
The call is fundamentally rooted in the fact that artificial intelligence companies audit the safety of the systems they develop themselves. Major tech companies signed a voluntary security agreement at the White House in September.
However, there is no enforcement mechanism requiring the implementation of these commitments. Experts argue that rules applied voluntarily by companies are insufficient.
Disagreements Among Experts
The scientists' call came after it was revealed that some advanced artificial intelligence models were able to bypass safety restrictions in test environments. Such incidents serve as a warning for potential risks in the future.
Conversely, Yann LeCun, who won the Turing Award alongside Hinton and Bengio, did not sign the letter. LeCun argues that software flaws, rather than artificial intelligence getting out of control, lie at the root of safety issues.
Future Uncertainties
Scientists stated that organizations should immediately adopt stricter safety standards until an intergovernmental joint oversight system is established. The European Union, on the other hand, argues that the current Artificial Intelligence Act possesses sufficient tools.
The call regarding the establishment of an international oversight body is not yet a governmental decision. How the proposal will be implemented and which countries will participate remain uncertain.