Summary

  • More than 70% of enterprises experienced AI-related breaches in the past year, with generative models becoming the primary target, says a recent report by the SANS Insititute.
  • At the same time, enterprise security investments focused on AI risks are growing significantly and there is a large gap in preparedness as AI attack surfaces rapidly expand, the report said.
  • State-sponsored attacks on AI infrastructure have spiked 218% year-on-year and CrowdStrike’s CEO, George Kurtz, has warned that ‘Security can’t be bolted on; it has to be intrinsic’
  • In a new development, CrowdStrike has embedded Falcon Cloud Security directly within NVIDIA’s universal LLM NIM, securing over 100,000 enterprise-scale LLM deployments across NVIDIA’s hybrid and multi-cloud environments.
  • With this embedded approach, CrowdStrike enables Falcon to continuously scan containerised AI models prior to deployment, proactively uncovering vulnerabilities, poisoned datasets, misconfigurations and unauthorised shadow AI.

By Louis Columbus

Original Article