Open models enable forensic analysis that commercial guardrails blocked
Original titleOpen models are more secure.
AISummary
A security team found commercial frontier model APIs blocked their incident-response log analysis, which required submitting real attack commands and exploit payloads. They ran the forensic analysis on GLM 5.2, an open-weight model, on their own infrastructure, which also kept attacker data and referenced credentials inside their environment.
Source: Bryan Catanzaro · x.comPublished · added here