Many companies leverage AI for innovation or automation, without considering risks. While any business activity or tool has risk, AI development projects involve unique risks and considerations.
Employee Risks
Your employees are using AI, whether you know it or not. A Deloitte study characterizes that over 60% of employees use AI at work. Yet well under 50% of companies have an AI acceptable use policy. Fewer still have AI Awareness training.
Several companies have seen their intellectual property compromised as it’s put into the public domain via employee use. Other companies have seen factually incorrect information (AI hallucinations) advanced in company output.
If you don’t define the rules via policies and train your employees, the liability is on you, the business.
Financial and Reputational Risks
Misuse or mistakes from AI fall back on the company. This may undermine your reputation, or may cost you directly.
In several instances, bad actors (or simply mischievous ones) have tricked AI chat bots into making promises on behalf of the company. In at least one case, a court required the company to honor the promise.
When your employees, or even your AI bots, make mistakes. The company is responsible.
Training Data Risks
Model Risks
Output Risks
Sensitive Data
PHI, PII, Biometrics
Racial/Ethnic, Financial, Educational, Location
Training Data Provenence
Ownership, Usage Rights/Consent
Source, Accuracy, Integrity, Bias
Bias Risks
Employment, Financial, Justice/Legal
Decision Risks
Critical Infrastructure, Healthcare, Safety
Liability & Reputational Risks Hallucinations, Intellectual Property rights, Privacy/Breaches, Unintended Consequences
Misuse of Output Data
Deepfakes, Propaganda, Fraud
AI-Specific Cyber Security Vulnerabilities
Model Inversion, Model Extraction, Model poisoning, Adversarial Attacks