Cisco sees great potential for artificial intelligence (AI) across its networking and security stack. It also recognizes the rapidly evolving technology is strewn with ethical pitfalls. Cisco’s Responsible AI initiative, announced today, is how the networking giant plans to navigate that minefield.

"As AI has become increasingly integrated throughout the products end-users interact with every day, questions of fairness, transparency, and privacy have become salient for AI developers and retailers," Anurag Dhingra, CTO of collaboration at Cisco, told SDxCentral in an email.

Cisco reports more than half of customers surveyed in a recent data privacy report, expressed concern with how AI is being used today, and that they are less likely to trust companies that utilize AI to determine credit worthiness, getting a job interview, or setting prices for products or services.

Cisco enacts AI oversight

Over the past few years, Cisco has explored several opportunities for AI across its vast software and hardware portfolio, ranging from machine learning-based security capabilities to AI operations capabilities for SD-WAN and LAN deployments. Many of these capabilities are predicated on access to large amounts of data, something Cisco has leaned heavily on ThousandEyes and AppDynamics.

For now, these data-driven insights have been limited to raising alerts and recommending configuration changes. However, the company has ambitions to close the loop on its AI capabilities and, in some cases, cut the human element out entirely.

One example is the integration of ThousandEyes network performance monitoring capabilities into Cisco's Viptela SD-WAN platform to provide real-time insights into WAN performance and to inform routing policy. The company aims to automate this process to dynamically deploy and route traffic across multiple ephemeral WAN links based on network conditions.

However, data privacy, particularly around AI, remains a top concern among customers, according to Cisco's Data Privacy Benchmark study.

Cisco's AI initiative aims to address many of these concerns by enacting checks and balances for the implementation of AI across the company’s product stack.

A core component of this initiative is oversight of AI applications and will see the vendor establish an AI committee made up of executives from its various business units.

The committee won’t, however, include members from outside the company, but “may reach out and rely on external stakeholders for guidance and alignment to market, standards, regulations, and industry cooperation when needed,” Dhingra said.

The committee will be responsible for advising on AI practices and overseeing the adoption of Cisco's broader responsibility framework. The committee will also review high-risk AI use cases proposed by business units and investigate reports of bias or discrimination in training or inference models.

Keeping bias and discrimination out of AI models is a major consideration and one that has dogged tech giants, including cloud giants like Google.

“AI is inherently different from previous technologies and requires a more responsible approach to governance,” Dhingra wrote. “AI models are typically trained on data sets and automate the product of insights that can influence decisions and actions.”

However, if that data is inconsistent or based on incomplete data sets, it can introduce bias, he explained.

Cisco approaches AI with caution

To mitigate this risk, Cisco’s Responsible AI initiative calls for strict controls and incident management to ensure the company’s security, privacy, and human rights policies are reflected in its AI design.

Cisco has also enacted incident management protocols for investigating cases of AI bias or discrimination which could have a negative impact on users. If bias is identified it's assigned and reported to the appropriate team.

"These controls are not simply guidance. We lead with them as a by-design development requirement to ensure that privacy, security, and human rights are built in from inception," Dhingra told SDxCentral.

Finally, Cisco is working with industry leaders, governments, and educational institutions to establish best practices and further the development of responsible AI models.