Most companies admit their current security can't stop AI cybercrime (www.techradar.com)

A Lenovo survey of 600 global IT leaders found a clear lack of confidence in today’s cybersecurity stack: 65% said their defenses are outdated and unable to stop AI-powered attacks. Respondents flagged three top worries — external threats, insider risks and defending AI itself — noting that generative AI makes attacks “faster, more convincing, and harder to detect,” with examples like polymorphic malware, AI-driven phishing and deepfake impersonation. Almost 70% worry about employees misusing AI, and over 60% say autonomous AI agents create a new class of insider threat their organizations can’t currently manage. Lenovo urges “fighting AI with AI,” proposing a two‑pronged strategy to harden detection and embed adaptive AI into existing security controls. That implies deploying continuous anomaly detection, model-integrity monitoring, tighter access controls around models/training data/prompts, and AI-driven SOC tooling that can keep pace with polymorphic and social-engineering attacks. Adoption faces real obstacles — legacy systems, talent shortages and budgets — but Lenovo argues that securing AI is not just defensive: protecting AI workloads is becoming a competitive differentiator that unlocks productivity while reducing risk. The takeaway for the AI/ML community is clear: defenders must prioritize adaptive, model-aware security controls as models themselves become high‑value assets and attack surfaces.