Tricentis, a software testing company, has confirmed that its security team identified and addressed a prompt injection vulnerability in an AI-powered capability during pre-release red-teaming. The flaw led to a one-week delay in the feature's deployment while backend fixes were implemented to prevent potential data exposure.
Erika Dean, CISO at Tricentis, emphasized the company's commitment to preventing the use of production data in non-production environments. She noted that while historically challenging for specific testing scenarios like load testing for check images or Social Security Number validation, current technology offers viable alternatives. This stance applies even to highly regulated sectors such as financial services and healthcare, where production data is sometimes used in testing despite weaker controls compared to live systems.
Dean also detailed Tricentis' rigorous evaluation process for AI model providers and vendors. Key disqualifiers include a lack of clear answers regarding data residency, retention policies, and whether customer data is used for model training. Vendors that cannot transparently explain their architectural approach, particularly when wrapping other models, are also rejected due to concerns about data control.
For smaller security teams, Dean recommended prioritizing three core areas: establishing a vulnerability management program for scanning and identifying security flaws, implementing monitoring solutions (either in-house or outsourced) to detect malicious activity, and focusing on corporate security fundamentals such as endpoint encryption, antivirus/malware detection, patching, and data loss prevention. She highlighted AI's potential to assist smaller teams in triaging alerts and developing new detection rules.
Tricentis' internal security strategy also involves automating evidence gathering and audits for compliance, which frees up resources to focus on enterprise and product security. This approach allows the security team to concentrate on deploying and managing technical controls to protect both the company and its customers, particularly as threat actors increasingly leverage AI to exploit vulnerabilities.
