Jaipur, March 21 A neuro-engineering deep-tech AI company based in Gujarat has developed an "AI Action Firewall" aimed at making artificial intelligence systems safer by verifying every action before execution, an official said on Saturday.
The firewall will function as a policy-based safety layer between AI systems and real-world operations, ensuring that actions are authorized, monitored, and recorded, he said.
A firewall is a network security system for monitoring and controlling incoming and outgoing network traffic based on predetermined security rules.
"The concept draws inspiration from traditional firewalls, but is specifically designed to regulate AI-driven actions such as sending emails, executing code, accessing databases, or triggering automated workflows," Saurabh Patel, Founder and CEO of NeuroPause Lab Limited, told reporters here.
Patel said his company applied for a global patent for the technology last year and is currently in the process.
Under the model, each AI action will be evaluated against predefined policies and classified as "allow", "review", or "block".
While routine actions could proceed automatically, higher-risk operations would require human approval, and unauthorized actions would be stopped, Patel said.
He added that all decisions will be logged to create a permanent audit trail, helping organizations ensure transparency, accountability, and regulatory compliance.
He noted that existing cybersecurity systems are designed for human users and cannot manage autonomous AI agents, making a dedicated control layer necessary.