Security risks specific to AI

AI Security and Risk Management

Angeline Corvaglia

Founder and Digital Transformation Specialist

The importance of risk management

• Harness benefits safely and responsibly
• Approach to security must change
• Traditional security strategies aren't sufficient

Think of the impacts of integrating AI as a coin:

• Potential
• Security risks

Subtle yet dangerous risks

• Bias
• Hallucination
• Data poisoning
• Adversarial attacks
• Sensitive data exposure

AI bias in hiring system

1. Training the AI with historical data
2. Reflect past biases
3. Magnifying the bias
4. Declining trust in the organization

Manipulating an online shopping platform

1. Platform with AI recommendation system
2. Fake data fed in
3. AI recommends unpopular or harmful products
4. Numerous potential consequences

AI bias

Lightning-fast decisions, amplify the impact of bias

Unfairly favor or harm

Uneven outcomes at an alarming speed

Hallucination

• Presents incorrect or misleading information as truth

Can lead to:

• Flawed decision-making
• Misinformation and lack of trust

Data poisoning

Intentional tampering with the training data

Causing the AI system to veer off course

• Incorrect predictions or decisions
• Significant consequences

Adversarial attacks

Presenting misleading inputs

• During its operation
• Challenging to detect
• Compromise integrity

Sensitive data exposure

Large amounts of data for training and operation.

Risk of exposure of sensitive data if not adequately secured

Understand risks to be able to manage them

Proactive risk mitigation

• Anticipate risks
• Know the AI security ecosystem
• Future-proof AI security
• Build a security-aware culture
• Understand model and external risks
• Align AI security with organizational goals

Let's dive in!

AI Security and Risk Management