Major AI developers should provide legal and technical safe harbors to protect public interest safety research from account suspensions or legal reprisal.
