At Sea AI Lab, we are dedicated to advancing Trustworthy AI by ensuring the safety, privacy, legality, and fairness of AI systems. Our research addresses these critical challenges, prioritizing human welfare and the responsible deployment of AI technologies.

Safety: Safety is a cornerstone of AI adoption. We develop rigorous methodologies, including adversarial testing and red teaming, to evaluate and enhance the reliability and resilience of AI models. Our work focuses on foundational models, such as large language models and vision-language models, ensuring they perform robustly in diverse environments and agent-based scenarios.

Privacy: Protecting user privacy is crucial in today’s data-driven world. Our research explores innovative privacy-preserving techniques, emphasizing practical approaches to assessing and strengthening data privacy guarantees. We aim to develop AI systems that uphold data security and user trust while maintaining performance and usability.

Legality: As AI adoption grows, navigating legal and regulatory frameworks becomes increasingly important. Our research examines key legal concerns, including copyright protection, data ownership, and intellectual property rights. We also develop data attribution and machine unlearning techniques to mitigate legal risks, ensuring compliance, transparency, and ethical AI usage.

Fairness: AI systems must be equitable and unbiased to foster trust and inclusivity. Our research identifies and mitigates biases in AI algorithms, promoting fairness across diverse demographic groups. By developing robust evaluation frameworks and bias-mitigation techniques, we strive to create AI that is fair, transparent, and aligned with societal values.