The General Incorporated Association Cybersecurity Alliance has published four key points for safe AI use based on its survey findings. The findings highlight practical conditions and challenges, including differences by company size and concerns about information leakage.
The 2026 Survey of Actual Security Practices was conducted among 500 people responsible for website operations and administration, including at global companies, to analyse the current state of and challenges in AI security. Establishing AI usage rules was the most common measure, at 38.6%, while verification against AI-specific attacks stood at 20.6%, application vulnerability assessments at 26.6% and the implementation of access restrictions at 18.0%, indicating that the proportion of technical verification measures was relatively low.
By company size, the implementation rate for AI usage rules was high at 50.9% among companies with 1,000 or more employees. However, verification against AI-specific attacks and access-rights configuration each stood at only 25.7%, leading to the observation that security cannot be assessed based on company size alone. The survey identified it as important to combine rules and technical verification suited to the AI systems in use.