Study finds trust in automated AI vulnerability scanning dropped to 9% (infosecurity-magazine.com)

• Organizations relying solely on AI for testing fell from 29% to 9%.
• 47% now prefer a hybrid human-AI model.
• 78% report automated scanners miss critical vulnerabilities.

"A Cobalt study based on surveys of approximately 450 cybersecurity professionals reveals a significant drop in trust for fully automated AI vulnerability scanning. The percentage of organizations relying solely on AI automation for testing fell from 29% to 9% over one year. Nearly half (47%) now prefer a hybrid model combining human oversight. 78% reported that fully automated scanners miss critical vulnerabilities. The study highlights the increasing complexity of the AI attack surface, with one in three AI pentest findings rated high risk."