Why does this keep happening?
Three weeks, three disclosures, and two of them trace back to the same outside evaluator accidentally wiring up internet access. That's a lab-safety story more than an alignment one, and it raises the question of who audits the auditors.
Three weeks, three disclosures, and two of them trace back to the same outside evaluator accidentally wiring up internet access. That's a lab-safety story more than an alignment one, and it raises the question of who audits the auditors.