AI isn’t ready to go solo: Expert urges collaborative security strategy

Fully autonomous artificial intelligence systems are already showing significant flaws, highlighting a critical need for human-AI collaboration in cybersecurity, according to Radiant Logic CEO Dr. John Pritchard.

Speaking at Identiverse 2025, Pritchard argued that relying solely on AI for identity security overlooks essential human capabilities such as adaptability, creativity, and nuanced judgment. Instead, he advocated a shift toward what he termed "multiplayer AI”. This approach, he said, emphasizes seamless coordination between specialized AI tools and human expertise.

"It's not about building the AI super soldier. It's about getting humans and AI to work together," Pritchard said during his keynote titled "Ready Player Two: Why Multiplayer AI Beats Going Solo in Identity Security." He likened today’s cybersecurity environment to chaotic virtual battlefields depicted in popular video games like "Ready Player One," emphasizing that contemporary threats mirror these unpredictable and fast-evolving scenarios.

"Our threats today are going to behave exactly the same way," he noted, suggesting that isolated defensive strategies are no longer sufficient.

To illustrate the limitations of autonomous AI, Pritchard recounted a significant event from 2018, when OpenAI’s highly trained bots lost to human players in the game Dota 2. Despite extensive training and significant resources, the AI bots couldn't swiftly adapt to unpredictable human strategies. "AI is really, really good at reacting to things it's been trained on," he explained. "But when it encounters something unfamiliar, it falls down."

He further underscored the necessity of improved human-AI coordination by referencing a real-world cybersecurity assessment conducted by the Cybersecurity and Infrastructure Security Agency (CISA) in 2023. In that scenario, ineffective communication between an organization’s Solaris IT team and its Security Operations Center allowed a simulated nation-state attack to persist unnoticed for months.

Pritchard cited this example to illustrate how disjointed communication and siloed teams compound risk, highlighting precisely why integrated human-AI collaboration is essential.

Pritchard also pointed out systemic challenges within identity security itself. Referencing insights from a CyberRisk TV wrap-up at Identiverse 2025, he noted, “Identity-related attacks are now the dominant threat vector in cybersecurity, yet most organizations remain hindered by fragmented tools, siloed data, and disconnected teams.”

This observation reinforced his message that without coordinated workflows, both human and AI defenses lose relevance and effectiveness.

Additionally, Pritchard detailed the concept of agentic AI, describing it as a model designed for collaborative problem-solving. He explained that agentic AI involves multiple specialized AI agents working collectively alongside human teams, each capable of communicating, sharing experiences, and delegating tasks. Such systems enhance overall operational resilience by leveraging human insight alongside AI efficiency and speed.

Moreover, Pritchard emphasized the significance of new open standards, including the Model Context Protocol (MCP) and Agent-to-Agent (A2A) interoperability. These standards enable diverse AI systems from different vendors to communicate effectively, share insights, and collectively address complex cybersecurity challenges, thus enhancing organizational agility and responsiveness.

Referencing Gartner research, Pritchard highlighted that teams combining human expertise with AI capabilities consistently outperform groups composed solely of either humans or autonomous AI agents. Specifically, Gartner's findings indicate that teams leveraging both human and AI collaboration demonstrate superior adaptability and decision-making abilities in complex, rapidly changing cybersecurity environments compared to single-dimension approaches.

Pritchard urged cybersecurity professionals to rethink their strategies around cooperative, multiplayer-style defense mechanisms. Echoing his initial message, he emphasized, "Let's stop grinding alone. Let’s embrace this multiplayer world. Let's team up and win this together," reinforcing the idea that successful cybersecurity requires both human ingenuity and AI collaboration.