This is the only primary-source interview on the METR report I have seen. This is one of the most important topics in AI this year, given the severity of the incident, we should all be listening to hear how we can put proper guardrails to ensure agents are securely and reliably deployed.