Executives warned that risks such as prompt injection, data exposure and vulnerabilities in widely used models could trigger large-scale impact, raising the threat of a wave of shadow AI-or use of unapproved AI tools-if governance frameworks fail to keep pace.