Call for an international auditing mechanism for artificial intelligence models
Experts from 34 different countries are demanding binding and independent audits for the security of advanced artificial intelligence technologies.
Scientists from 34 countries around the world have published an open letter calling for the auditing of advanced artificial intelligence systems and the enhancement of security.
Demand for an international auditing mechanism
In an open letter signed by scientists from 34 countries, it was proposed to establish an international auditing mechanism to ensure the security of advanced artificial intelligence systems. It was requested that the proposed organization establish safety and control standards to be applied across the industry.
Registry system and transparency expectation
The letter stated that a registry system recording major safety incidents should be established, and these records should also be shared with the public in the future.
Isolated environments and mandatory independent audits
It was stated that the most advanced artificial intelligence models should initially be operated in environments completely isolated from external networks and be subjected to independent audits.
Insufficiency of voluntary commitments
Arguing that the security commitments voluntarily undertaken by companies are insufficient, the researchers emphasized that official steps must be taken.
White House agreement and signatories
This call came about a month after leading technology companies signed an agreement with the White House accepting voluntary security measures. Among the signatories of the letter were Geoffrey Hinton and Yoshua Bengio, while Yann LeCun did not sign the letter.