35% of Generative AI Users Experience Near-Miss Incidents Due to Overconfidence and Data Input

Cyber Security Cloud conducted a survey of 300 office workers on generative AI usage. The study found that while 67% use AI for work, 15% operate as 'Shadow AI' without company approval. Approximately 60% of users have input internal data, including financial and customer information. 35% reported near-miss incidents, with 14% resulting in actual problems. The primary causes are identified as overconfidence in AI outputs and lack of verification, highlighting the urgent need for corporate governance and employee education.