67% of companies deploy AI without security testing

Most companies using AI tools do not assess their vulnerabilities before implementation, which leads to data leaks and cyberattacks. According to Informzashchita , 80% of organizations have already experienced information compromise due to unsafe use of AI.

Tech May 21, 2025 0 32 Add to Reading List

67% of companies deploy AI without security testing

67% of companies implementing AI tools do not assess their security and impact on infrastructure security in advance. According to Informzashita, the priority for businesses remains the speed of implementation, and security issues are resolved post factum, when AI has already been integrated into the system. The most common incident is data leakage — 80% of organizations have encountered it. Without a preliminary risk assessment, an AI solution can gain access to excessive confidential information, transmit it through vulnerable APIs, or store it in unreliable databases, which increases the risk of hacking and data loss.

Another common threat is prompt injection attacks, where attackers manipulate requests to AI to change its behavior. This allows them to obtain sensitive data, inject malware, or change security settings. More than 40% of companies that skipped the AI tool validation stage faced such attacks.

Small and medium businesses are particularly vulnerable: they account for 90% of attacks through AI vulnerabilities. These companies often implement technologies in a hurry, trying to optimize processes or reduce staff, and neglect data protection. Experts advise conducting information security audits and pentests before integrating AI solutions to minimize risks.