What Are the Security Concerns Around AI-Driven Networks?

_{WHAT ARE THE SECURITY CONCERNS AROUND AI-DRIVEN NETWORKS?}

^{Data Privacy Tops the List of Security Worries}

What is your organization's biggest security concern with AI-driven networks?

Would you trust AI to automatically quarantine a major part of your network without human review if it detected a threat?

_{WHAT ARE THE SECURITY CONCERNS AROUND AI-DRIVEN NETWORKS?}

Data Privacy and AI training data exposure is the biggest security worry

68.8%

would trust AI to automatically quarantine their network

The second biggest security worry was AI making incorrect security decisions

The biggest security worry with AI-driven networks is data privacy and AI training data exposure, with 47.1% citing this as a concern. This highlights the need to introduce stringent governance frameworks and robust access controls to limit the data that AI networking models have access to, and to limit human access to the models too.

The second biggest concern was AI making incorrect security decisions (39.4%). Interestingly, though, 68.8% said they would trust AI to automatically quarantine a major part of their network without human review if it detected a threat, so long as a confidence threshold is met.

Perhaps this reflects respondents’ confidence in their organization’s ability to implement robust threshold architectures and calibrate them appropriately to mitigate the risk of poor AI decision-making.

Answers were more nuanced when respondents were asked why they answered as they did. The importance of human oversight in AI decisions, particularly when critical systems and data are involved, was mentioned often. Several commented that it’s simply too early for enterprises to be able to trust autonomous AI without human oversight.

_{WHAT ARE THE SECURITY CONCERNS AROUND AI-DRIVEN NETWORKS?}

^{Enterprises Are Cautiously Optimistic About Security Risks in an AI World}

Do you believe AI-driven defenses will outpace AI-driven attacks in the next five years?

When it comes to security, 37.2% of respondents said that AI-powered attacks evolving faster than defenses was a concern with AI-driven networks.

Nearly two-thirds (65.3%) of people in our survey, on the other hand, said they believe defenders will outpace attackers in the AI-driven security arms race. A further 23.5% believe attacks and defenders will remain in a constant arms race.

These results seem to reflect a wary optimism: while AI-powered threats pose a real and alarming danger, enterprises believe that ultimately AI-powered security systems will prove to be more robust than the tools used by attackers. It’s also possible that respondents simply see AI-driven networking as more vulnerable than other enterprise AI use cases.

Some respondents pointed out that while the rapid and continued evolution of AI-generated threats is inevitable, security vendors and service providers are pouring huge resources into designing defenses against these attacks. Others mentioned that this kind of arms race is simply ‘business as usual,’ and that historically the majority of enterprises have successfully defended against attacks.

_{WHAT ARE THE SECURITY CONCERNS AROUND AI-DRIVEN NETWORKS?}

13.4%

more vulnerable

Do you believe AI will make networks MORE secure or MORE vulnerable in the next 5 years?

Do you believe AI-driven networks require a fundamentally different security architecture than traditional networks?

On average, respondents believe that AI will make networks 13.4% more vulnerable in the next five years.

AI has enormous potential to help enterprises strengthen their security posture, by speeding up the detection and mitigation of vulnerabilities, threats, and breaches, by prioritizing alerts, and by analyzing huge amounts of security data for valuable insights. However, incorporating AI into the enterprise network can introduce new security risks and vulnerabilities that could be exploited by attackers.

So it’s not surprising that 59.5% of respondents believe AI-driven networks require a fundamentally different security approach compared to traditional networks. Legacy security approaches are too static and rules-based for a dynamic AI-driven network architecture. Instead, ‘never trust, always verify’ approaches like zero trust will be critical to secure AI-driven networks, along with stringent governance, and effective incident response plans.

_{WHAT ARE THE SECURITY CONCERNS AROUND AI-DRIVEN NETWORKS?}

What's your biggest trust gap in AI security?

The two biggest trust gaps in AI security are managed AI providers which aggregate multi-client data (33.8%) and AI vendors and their security practices (31.4%). Given this, managed AI providers that can demonstrate they use customer data in a strictly compliant and anonymized way (or that already have a sufficient volume of data and therefore don’t need to collect more) will have an advantage.

When asked what effect this trust gap was having on their organization, some said that it was slowing their adoption of AI, that they were having to be more conservative in its use, or to put a lot of resources into due diligence, compliance, and security. Others noted that this was making them fall behind in the race to harness AI.

Another common theme was a lack of internal alignment on the best approach to AI, slowing decision-making, and creating conflict. Several said the trust gap meant they were avoiding external vendors altogether, or were taking longer to find partners they could trust.