AI Experts: Secure Data Essential in AI Development, Deployment

Artificial intelligence experts are refocusing their efforts on security, protecting the emerging technology’s three weak spots: algorithms, hardware and training data.

The AI supply chain is reliant on data, including critical and classified information. Due to data’s vital role, it is often the target of malicious activity, usually tampering with inputs added to the AI model, Signal reported.

Cloudera Government Solutions President Robert J. Carey, a past Wash100 Award winner, said data integrity is important in implementing any decision-making solutions.

To address the vulnerability of training data, Darktrace Federal CEO Marcus Fowler said unsupervised learning is a solution to reduce malicious interference. Development, security and operations benefit the software production chain from using the approach.

Fowler added that complying with the U.S. regulations on the origin of critical components and supply chain traceability rules is another aspect that will significantly improve the security of data used in AI models.

AI security is a must for the U.S., as it is planning to implement AI to address cyberthreats. Former National Security Agency Director Michael Rogers said the technology can help in identifying attackers’ strategies, apart from improving readiness.