AI has captivated widespread interest due to its numerous benefits. However, its rapid advancement and widespread adoption raise concerns, particularly in cybersecurity. The influx of insecure applications on various devices and endpoints opens more pathways for cybercriminals to steal data.

Open-Source AI Security Challenges

Open-source AI applications face significant security challenges. These applications are often developed and supported by volunteers, making them vulnerable due to the lack of dedicated security resources. Despite their benefits, the widespread availability of open-source AI projects increases their risk of compromise.

AI is Software

It’s crucial to remember that AI is fundamentally software and part of the software supply chain. AI security should be integrated into the broader context of software security, encompassing every stage from development to deployment and maintenance.

Risks in the AI Software Supply Chain

The AI supply chain faces risks similar to the broader software supply chain but with additional complexities. For example, a financial institution using AI for loan risk assessment must ensure the AI model and its training data comply with regulatory standards to avoid biases and legal issues.

Security Concerns

The popularity of open-source AI tools inversely correlates with their security posture, meaning widely adopted tools often have more vulnerabilities. Furthermore, AI models trained on potentially unethical data pose significant legal risks, highlighting the need for stringent security measures.

Steps for Security Professionals

Security professionals can improve open-source AI security through several avenues:

• Security Specifications: Advocate for transparency and accountability within the open-source community, demanding security metadata like Software Bill of Materials (SBOMs), SLSA (Supply Chain Levels for Software Artifacts), and SARIF (Static Analysis Results Interchange Format).
• Open-Source Security Tools: Collaborate with companies that support security projects, such as Allstar, GUAC, and in-toto attestations, to benefit from open-source innovation while managing liability.
• Industry Contributions and Funding: Support organizations like the Open Source Security Foundation (OpenSSF) to secure critical open-source projects through specifications, tools, and initiatives.

CISOs and their security teams need detailed information about the software in their environments to make informed, risk-based decisions. Relying solely on volunteer efforts for security is unsustainable and ineffective.

Conclusion

Securing open-source AI is crucial for its responsible adoption and sustained success. By enhancing security measures and supporting open-source security initiatives, the cybersecurity community can address these challenges and protect against vulnerabilities.