A visitor poses for pictures with a humanoid robot at the 2025 Global Artificial Intelligence Machines and Electronics Expo in Macao, South China, Dec 4, 2025. [Photo/Xinhua]

As artificial intelligence accelerates both digital transformation and cyber threats, enterprises must move from passive, reactive security models to proactive and autonomous defense systems to keep pace with increasingly sophisticated attacks, a senior cybersecurity executive said.

According to the latest forecasts by cybersecurity company Palo Alto Networks, 2026 will mark the "year of the defender", as autonomous AI-driven security becomes essential to counter AI-enabled identity attacks, data poisoning and emerging quantum risks, said Adrian Chan, vice-president of Palo Alto Networks Greater China.

"As enterprises adopt AI, attackers are doing the same,"Chan said. "Once hackers discover a vulnerability, they can use generative AI to rapidly create attack tools, generate variants and launch large-scale attacks. In many cases, before companies are able to deploy patches, systems may already be compromised."

The speed of cyberattacks has increased dramatically in recent years, he noted. In 2021, it took an average of nine days to form a full attack chain. By 2023, that time had shortened to about two days. "In 2025, there are cases where attacks were executed within 30 minutes," the senior executive said.

"This shows that traditional, passive defense methods are no longer sufficient in the AI era," Chan said, warning that attackers are increasingly leveraging AI to scale up threats across a hybrid workforce, where autonomous agents already outnumber humans by an estimated ratio of 82 to 1.

"This requires a fundamental shift — from being a reactive blocker to becoming a proactive enabler that manages AI-driven risks while supporting enterprise innovation," Chan said.

The company also forecast that by 2026, about 40 percent of enterprises will deploy AI agents in their operations. However, only 6 percent currently factor security into the design of those agents.

"One way out is to 'secure AI by design'," Chan said. "Security must be built in at the very beginning of AI system design, rather than added later."

Based on a survey of 1,000 large enterprises worldwide, Palo Alto Networks found that companies use an average of 85 security tools from 29 different vendors. Such fragmented systems often operate in isolation, creating blind spots and increasing vulnerability.

"In the AI era, this fragmented and reactive approach is simply too slow," Chan said. "With so many disconnected tools, every step could be a potential vulnerability. Once a flaw is found, attackers can immediately exploit it using generative AI."

In this regard, a platform-based security approach that integrates network, cloud and endpoint protection into a unified system may be a rather efficient way to address the challenge, Chan said.

"Through platformization and AI-driven analytics, enterprises can gain real-time visibility into attack paths, detect threats as they occur and deploy coordinated defense strategies across environments," he added.