Tencent Cloud Launches Thorough Framework for AI Risk Management
Tencent Cloud has introduced a new framework designed to help enterprises assess adn control risks associated wiht artificial intelligence (AI) systems, aiming to foster a more trusted AI ecosystem. The framework addresses the growing complexity and diversity of potential AI-related risks.
The framework categorizes AI system risks into six core modules: governance,data security,large model ontology,agents,tools & boundaries,and AI infrastructure. It further details these risks through seven levels, encompassing 26 control modules and over 130 specific control measures.This systematic approach is intended to provide robust security governance capabilities and has already been implemented internally within Tencent cloud to create end-to-end protection for its AI systems.
Recognizing the challenges enterprises face in implementing such security measures, Tencent Cloud has also launched a suite of security products built upon this framework, offering “out-of-the-box” solutions. these include:
* AI-SPM (big model situation management): This product focuses on AI security governance and observability, providing an attack surface and vulnerability management system for large models and agents. It aims to proactively identify and address security risks within the large model infrastructure.
* LLM-WAF (Large Model Firewall): Moving beyond traditional “static policy” approaches, LLM-WAF offers full-link protection across multiple models and scenarios, even in high-concurrency environments. It provides real-time detection and interception of threats like computing power abuse, prompt injection attacks, and data leakage.
* Tianyu (Big Model Security Gateway): This gateway centers on establishing a “trusted” identity and ecological system. It provides unified control over identity management, AI attacks, and the security of decision-making and execution within enterprise intelligent applications, creating layered defenses.
Beyond these core products, tencent Cloud also offers pre-online evaluation services and large-model security intelligence. the company emphasizes a holistic,end-to-end approach,integrating security capabilities throughout the entire AI lifecycle – from research and growth,through deployment,and into ongoing operation and maintainance.
tencent Cloud’s goal is to provide solutions that ensure AI systems are not only “capable” but also “reliable,” enabling enterprises to confidently leverage the power of AI.
