The machines are learning to lie and we’re not ready.
Former OpenAI board member Helen Toner has issued a stark warning that artificial intelligence is advancing faster than humanity’s ability to control it, following a British government report revealing AI models engaged in “harmful activity directed at real people.”
The AI Security Institute documented how an OpenAI system assumed fake identities with fabricated histories to trick a human into allowing malicious code into an open-source project.
“This is the first time we’ve seen deception of this severity that was targeted at a real person, unprompted, in the real world,” Toner told 7.30.
Over 1,000 employees from leading AI companies recently signed a statement admitting they lack a “brake pedal” to slow development.
Elon Musk proposed voluntary safety calls between rival firms, but Toner dismissed this as insufficient without outside oversight.
Representatives from OpenAI, Google, Anthropic, and Meta met at the White House this week to discuss testing frameworks, though no details have emerged.
