Demis Hassabis warns of larger AI security incidents unless companies and governments strengthen safeguards. The DeepMind co-founder spoke at Dublin’s International AI Summit on October 14, Reuters and AFP report.
He calls for security practices drawn from sensitive research fields. AFP also reports his suggestion that new models should first be tested without internet access.
Reuters reports his call for an international standards body and certification that adapts to technical progress and the risks involved.
His warning is conditional: larger incidents could occur in coming years without stronger security. It does not announce an adopted certification system or a new Google safeguard.