By purchasing Koi Security, a startup that specializes in shielding endpoints from the threats posed by sophisticated AI agents, Palo Alto Networks plans to expand its AI security toolkit. This final agreement, which was announced on February 17, 2026, aims to combat a growing threat: AI tools and agents that have extensive permissions and data access while eluding traditional security measures. These "agentic" systems pose a novel threat to endpoint security, necessitating creative countermeasures.

The goal of traditional endpoint security is to identify malware using behavioral abnormalities or signatures. AI agents, however, alter the rules. With little supervision, they actively read files, run scripts, and move data between systems.

These tools function as trusted insiders and gain access to confidential company data, such as customer records or intellectual property, without raising red flags from endpoint detection and response (EDR) platforms or antivirus software. Every laptop, server, and virtual machine has an unmanaged attack surface known as the "Agentic Endpoint," according to industry observers. This blind spot is already being exploited by attackers.

Through API-driven remote code execution (RCE), which enables malicious code to run undetected, or authentication bypasses, where stolen tokens grant unauthorized entry, they take advantage of weaknesses in agent frameworks. Techniques include injecting payloads through extensions, plugins, packages, scripts, or even AI model artifacts; spoofing agent identities to imitate authentic automation; and stealing credentials for lateral movement.

These threats integrate into regular developer workflows, making legacy controls obsolete, in contrast to traditional malware that was restricted to executables. Closing the Agentic Security Gap Palo Alto Networks intends to improve Cortex XDR for increased endpoint visibility and integrate Koi Security's technology into its Prisma AIRS platform, which already secures AI-driven operations. Real-time AI agent behavior monitoring, automated policy enforcement to limit dangerous behavior, and proactive malware prevention catered to agentic environments are all promised by this integration.

Dashboards displaying agent permissions, data flows, and anomaly detection unique to AI tools will be made available to organizations, bridging the gap between security rigor and development speed. AI agents, according to Palo Alto's Chief Product and Technology Officer Lee Klarich, are "ultimate insiders" with unrestricted access but no conventional barriers.

In the announcement, he said, "This acquisition gives customers the visibility and control to deploy agentic tools safely." This was echoed by Koi CEO Amit Assaraf, who pointed out that agentic setups are overlooked by traditional solutions. By collaborating with Palo Alto, Koi's innovations will be scaled to enterprise levels while safeguarding against exploits in tools such as those from Anthropic, OpenAI, or custom frameworks.

The agreement highlights a change in cybersecurity: endpoints become dynamic ecosystems that require agent-aware defenses as AI automates tasks like coding and customer support. This means that security teams should give runtime protection more importance than static scans. The acquisition, which is anticipated to close soon, puts Palo Alto in a leading position to secure the hidden risks of the AI era, even though the financial terms are still unknown. Make ZeroOwl your Google Preferred Source.