Former Deepmind PR staffer says the lab once banned public discussion of AI extinction risk
decoder ·
Vishal Maini, a former communications and policy staffer at Deepmind, revealed that the lab previously prohibited any public discussion regarding the risk of human extinction caused by artificial intelligence. During his tenure from 2018 to 2022, researchers were instructed to dismiss such concerns as alarmist comparisons to science fiction films like Terminator. Instead of addressing safety risks, staff were coached to pivot conversations toward positive applications in healthcare and climate change. Internally, however, the team acknowledged that the AI alignment problem remained unsolved and that insufficient personnel were dedicated to solving it. After months of internal resistance, Deepmind eventually relaxed these restrictions to allow positively framed safety content. This shift included publishing blog posts that discussed extinction risks using more accessible and friendly language. Maini notes that the disconnect between internal awareness and public messaging is narrowing due to increasingly undeniable evidence of potential dangers. His account aligns with a broader trend of AI safety researchers speaking out about unsecured models and systems that may exceed human control.
decoder