Artificial Intelligence (AI) Risks

Based on industry, security reports, and ethical perspectives, many security risks for using AI were identified and below are the top 10 AI risks:

1. Prompt Injection Attacks: Attackers may feed malicious instructions into AI inputs, which could override or bypassing security measures and control the behavior of the AI model.
2. Data Poisoning: Attackers may tamper the training data, causing the AI models to learn incorrect behaviors or create backdoors.
3. Insecure Output Handling & Data Privacy: AI-generated outputs may produce incorrect information.  If outputs are used without validation, it may allow attackers to achieve full control of the system and steal sensitive data via the system.
4. Deepfakes & Misinformation: AI can create convincing fake content (images, video, text), eroding trust and enabling social engineering attack.
5. Excessive Permission for Agency: Over-granted permissions to the AI system allows it to take actions (e.g., sending emails, making purchases) without human oversight, leading to unintended, potentially damaging outcomes.
6. Intellectual Property Infringement: Unauthorized access to copyrighted data to train the AI models and generate outputs which mimics creator’s work.
7. Bias and Discrimination: AI models consist of societal prejudices in their training data, leading to biased decisions or unfair outcomes which affect in hiring, lending, healthcare, or law enforcement, etc.
8. Cybersecurity Threats: AI tools enhance the capabilities of attackers to create sophisticated phishing, malware, or automated attacks.
9. Lack of Transparency and Accountability: The “Black box” AI systems make it difficult to explain the decisions, and difficult to determine the responsibilities when errors occur.
10. Economic Impact: Rapid automation of tasks causes workforce disruption and requires reskilling efforts.

Reference:

• PCPD: AI Security (AI Security)

Published on:  Apr 2026