← Back to feed
general Cyberscoop

Irregular says ‘human oversight’ responsible for AI sandbox escape incidents

Cyberscoop

In a post-mortem, the frontier AI testing company said internet access for models is necessary to fully test out their cybersecurity capabilities. The post I...

Read the full story Cyberscoop →

Related Coverage

general Cloud Range lets SOCs benchmark AI agents against human defenders Help Net Security · Sep 24 general OpenAI agent breached Australian government health website, Albanese says The Record · Sep 24 general Gurucul connects AI activity to identity data for faster threat response Help Net Security · Sep 24