AI 보안 침해 비용 $5.33M — 92%는 기본적인 잠금장치조차 없었다
IBM 보고에 따르면 AI 보안 사고를 당한 기업의 92%가 접근 통제를 갖추지 못해 침해 비용이 $5.33M에 달했으며, 보안에서는 모델 선택보다 기본 수칙이 더 중요하다는 것을 입증한다.
ibm's latest research with the ponemon institute across 602 companies found that 92% of organizations experiencing ai security breaches failed to implement adequate access controls. incidents involving ai cost an average of $5.33 million, significantly higher than the global average of $4.99 million, and peaked at $6.04 million when attackers themselves utilized ai tools. the study highlights that security failures typically stem from compromised apis, connected applications, or misconfigured cloud services rather than the ai models themselves, regardless of whether they are open-source or proprietary. this suggests that enterprises are prioritizing model adoption over foundational security protocols, creating expensive vulnerabilities.