Sama, a leader in AI data annotation and model validation, recently launched “Sama Red Team,” a new initiative aimed at enhancing the safety and reliability of generative AI and large language models [LLMs]. The Red Team, comprising machine learning engineers, applied scientists, and human-AI interaction designers, focuses on testing AI models for fairness, safety, and legal compliance. By proactively identifying potential vulnerabilities in AI models across various modalities, including text, image, and voice, the team provides developers with crucial insights to address these issues before they become public.
Duncan Curtis, SVP of AI Product and Technology at Sama, emphasized the importance of these tests not only for performance but for developing responsible AI. He noted that although generative AI models appear reliable, they still require rigorous checks to ensure they adhere to public safety and privacy laws.
Sama Red Team tests models on four key competencies: fairness, privacy, public safety and compliance.
In fairness testing, teams simulate real-world scenarios that may compromise a model’s built-in safeguards and result in offensive or discriminatory content.
For privacy testing, Sama experts craft prompts designed to make a model reveal sensitive data, such as Personal Identifiable Information [PII], passwords or proprietary information about the model itself.
In public safety testing, teams act as adversaries and mimic real-world threats to safety, including cyberattacks, security breaches or mass-casualty events.
For compliance testing, Sama Red Team simulates scenarios to test a model’s ability to detect and prevent unlawful activities such as copyright infringement or unlawful impersonation.
The Red Team’s process includes consulting with clients to understand specific model behavior, conducting initial assessments, and testing with refined or new prompts to uncover vulnerabilities. This approach allows for large-scale testing if necessary, supported by Sama’s extensive network of over 4,000 trained annotators.
In addition to Red Team, Sama offers other services through its GenAI suite, enhancing the model development lifecycle from data creation to ongoing evaluation. This includes creating and reviewing model prompts and responses, ensuring they meet various criteria like accuracy and coherence.
All Sama services utilize SamaHub™, a collaborative platform for workflow communication and project tracking, and are supported by SamaAssure™, ensuring high-quality outputs with a 98% acceptance rate. The integration of human assessments and proprietary algorithms through SamaIQ™ further aids in identifying and addressing model vulnerabilities proactively.
