CrowdStrike launches SafeMind, pitting an attack model against a defence model
The product is not either model but the loop between them. Every performance figure in the launch is CrowdStrike's own, measured against competitors it has not named.
AI Safety and Security
Alignment research, model evaluations, security incidents, jailbreaks and vulnerabilities affecting deployed artificial intelligence systems.
OpenAI says Astra is its first model to reach Critical cyber capability
The model scored perfectly on ExploitBench and chained its own zero-days into a browser sandbox escape. Its full cyber capabilities will not be available at launch.