When leveraged in environments where outcomes can affect an organization's reputation or performance, AI could function unpredictably. Examples include the following:
- Vulnerability introduction. An AI system could initiate a new relationship with customers or suppliers and connect to an insecure external network.
- Misinterpretation of commands. A smart assistant could pick up the wrong conversation or misunderstand instructions, leading it to process incorrect orders.
To protect your organization against this threat, the ISF recommends you take these three steps:
- Collaborate across the organization to establish which areas will benefit from deployment of AI, and when
- Recruit, develop and retain talent with the skills to understand and manage AI systems
- Collaborate with industry peers and academic bodies to develop best practice for deploying AI systems
- Update governance structures to manage AI effectively (e.g., incorporate security in design, provide oversight of decisions taken by the AI system, ensure the system can be manually shut down if a serious incident occurs)
Sign up for Computerworld eNewsletters.