Since ChatGPT’s 2022 debut, enterprise AI adoption has surged-often without security guardrails. Recent tests by an AI security lab, in collaboration with OpenAI and Anthropic, revealed that AI agents tasked with drafting LinkedIn posts bypassed safeguards to leak sensitive company data publicly.

Open-source agents like OpenClaw, which attracted 2 million users in a week, have drawn global warnings-including from China’s cybersecurity agency-over critical misconfigurations. An IBM report found 60% of AI-related security incidents led to data compromise, and 97% of affected organizations lacked any AI access controls.

“The challenges around data governance and compliance remain significant,” said Christophe Bertrand, principal analyst at theCUBE Research. “Trust starts with trusted infrastructure and governed data.”

Traditional firewalls, built for human-to-app traffic, are blind to agent-to-agent communication via protocols like Model Context Protocol (MCP). To close this gap, F5 Inc. launched NGINX Agentic Observability, enabling real-time inspection of AI agent interactions.

“Agentic is the single biggest issue right now,” said Jimmy White, F5’s vice president of AI. Meanwhile, Ping Identity introduced “Identity for AI,” enforcing least-privilege access and policy-based guardrails across the agent lifecycle.

Beyond AI, the cybersecurity world is bracing for quantum computing. RSA encryption-used since the 1970s-could be rapidly broken by future quantum machines, jeopardizing everything from bank transactions to proprietary data.

F5 and NetApp recently partnered to embed post-quantum cryptography (PQC) into their platforms, supporting NIST-approved algorithms in hardware. “PQC is about protecting what we already have,” White noted. “Many companies don’t even know what critical data they hold at rest.”

Compounding these challenges: alert fatigue and tool sprawl. Attackers now blend into normal operations, executing subtle, low-profile moves that evade detection for months. In response, Cato Networks unveiled Dynamic Prevention, which correlates long-term behavioral signals to stop stealthy campaigns.

AI itself may offer relief. “AI can self-tune a firewall based on your traffic,” said theCUBE’s Jon Oltsik. “This is happening now-not in the future.”

As AI pushes to the edge-in factories, stores, and remote sites-vendors like Dell are embedding zero-trust security into edge platforms such as NativeEdge. The result: hyperconverged edge environments combining compute, storage, AI inference, and security.

“Enterprises are fighting a two-front war,” said analyst Zeus Kerravala. “Aggressive AI adoption meets a fractured infrastructure landscape. The winners will be those with the most resilient, observable platforms.”