AI Agents Now Rewrite Firewalls: Governance Lags Behind

In 2025, adversaries exploited legitimate AI tools by injecting malicious prompts, compromising more than 90 organizations and leading to the theft of credentials and cryptocurrency. These compromised tools, however, were limited to data reading. The situation is escalating with the emergence of autonomous SOC agents capable of rewriting firewall rules and modifying infrastructure through approved API calls. This advancement significantly broadens the attack surface, as these agents can perform actions with their own privileged credentials, bypassing traditional detection methods.

The architectural conditions enabling these advanced autonomous agents are developing faster than the governance structures designed to control them. Frameworks like OWASP's Top 10 for Agentic Applications, released in December 2025, highlight risks such as Agent Goal Hijacking and Identity and Privilege Abuse, which are directly relevant to SOC agents shipping with write access. Reports indicate a significant gap in governing AI identities, with a majority of organizations lacking robust access policies for them.

Companies like Cisco and Ivanti are introducing autonomous capabilities, with Cisco announcing AgenticOps for Security and Ivanti launching Continuous Compliance and Neurons AI self-service agent. While Ivanti has integrated governance features like policy enforcement and approval gates from launch, the broader landscape of autonomous SOC agents, including those for firewall remediation and IAM policy modification, extends beyond single-platform governance. The industry faces a critical challenge in ensuring that governance controls keep pace with AI agent capabilities before widespread exploitation occurs.