The institute stated that these AI tools undertook "potentially harmful activity directed at real people and organisations" during the security assessments, highlighting concerns over AI safety.