What is AI Application Impact Assessment (AIAIA) ?
The Artificial Intelligence (AI) is now a major part of our daily lives, transforming how businesses and communities operate. As AI becomes more widespread, concerns are growing about issues such as privacy, bias, ethical dilemmas, and societal harms.
AIAIA is a risk management tool designed to assist organizations in identifying and mitigating potential risks to ensure responsible AI deployment. AIAIA enables organizations to evaluate the impact of an AI system on users and society throughout its lifecycle. To facilitate this process, HKSARG has published a guide for those unfamiliar with responsible AI practices.
Framework of AI Application Impact Assessment (AIAIA) ?
Security : OWASP Top 10 LLM
The OWASP Top 10 for Large Language Model (LLM) Applications is a consensus-based awareness document published by the Open Web Application Security Project (OWASP).
- LLM01 : Prompt Injection - Attackers manipulate LLMs via crafted inputs to bypass safeguards.
- LLM02 : Sensitive Information Disclosure - LLMs accidentally leak sensitive data from training sets or user inputs.
- LLM03 : Supply Chain - Compromised pre-trained models or datasets introduce backdoors.
- LLM04 : Data and Model Poisoning - Malicious training data corrupts model behavior.
- LLM05 : Improper Output Handling - Unsensitized LLM outputs enable code/SQL injection.
- LLM06 : Excessive Agency - Overprivileged LLMs perform dangerous autonomous actions.
- LLM07 : System Prompt Leakage - Internal prompts exposing API keys or rules are extracted.
- LLM08 : Vector and Embedding Weaknesses - RAG systems return poisoned or unauthorized data.
- LLM09 : Misinformation - LLMs generate plausible but false content.
- LLM10 : Unbounded Consumption - Resource-heavy queries cause financial/performance damage.
References: https://genai.owasp.org/llm-top-10/
Privacy : PDPO Ethical Development and Use of AI Guidelines
This Guidance, released by PCPD, outlines Data Stewardship Values and Ethical Principles, to ensure lawful, ethical AI development, fostering stakeholder trust.
Workflow of AI Application Impact Assessment (AIAIA)
Enquiry for AI Application Impact Assessment (AIAIA) Service
We will collect and use your personal information solely for the purpose of addressing your inquiry regarding our services. After contacting you, we will cease to retain your personal information, including your name, number, and email. We will ensure it’s deletion.