Generative AI tools are being adopted faster than any technology before them, often without security review. The risks are real but manageable with the right governance.
Know where AI is used
Start with an inventory: which teams use which AI tools, what data they share, and which internal applications embed large language models.
Protect sensitive data
Set clear rules on what may be shared with public AI services, prefer enterprise offerings with data-protection commitments, and use data-loss-prevention controls.
Test AI applications for new attack classes
Prompt injection, jailbreaks, insecure output handling and excessive tool permissions are now among the top risks for LLM applications (see the OWASP Top 10 for LLM Applications). Test before launch and monitor afterwards.
Prepare people for AI-powered fraud
Deepfake voice and video are being used to impersonate executives and authorise payments. Introduce verification procedures for payment changes and urgent requests, and train finance and executive teams.
Govern with a recognised framework
The NIST AI Risk Management Framework and ISO/IEC 42001 provide structure for AI governance, risk and accountability.
Explore our AI Security service.