The OpenAI squad in charge of mitigating the risks of super-intelligent AI has lost nearly half its members, says a former researcher

• OpenAI initially had about 30 people working on AI safety, says ex-researcher Daniel Kokotajlo.
• But Kokotajlo says multiple departures have since reduced its ranks to about 16 members.
• Kokotajlo said people working on "AGI safety and preparedness are being increasingly marginalized."

OpenAI has lost nearly half of the company's team working on AI safety, according to one of its former governance researchers, Daniel Kokotajlo.