ACCORDING to Infosecurity Magazine, the rush to deploy AI in enterprise environments is not only creating new security vulnerabilities but reviving old security failures, with Jurgen Kutscher, VP of Mandiant Consulting, part of Google Cloud, warning of neglecting basic security controls. He said that “a lot of the old problems are new again” and that enterprises are worrying about AI threats like large language model poisoning while forgetting essential hygiene.
Mandiant’s red team has found real security failings during simulated attacks, including AI-enabled environments where an attacker could change data classifications to bypass protections such as data loss prevention. They also observed unencrypted communication streams and scenarios where social engineering provided initial access, after which the AI performed follow-on actions including exfiltration and policy changes.
The executive urged organisations to implement AI security governance, revisit secure architecture, and conduct red-team validation to ensure critical assets remain truly segmented.