Behavioral security for AI agents — local, private, and hash-only by design.
This platform and the detection methods, behavioral intelligence, scoring systems, and architectural patterns it embodies are covered by 223 claim families filed across two provisional patent applications with the United States Patent and Trademark Office (USPTO): the first on March 25, 2026, and a second, covering additional claims, on August 11, 2026. A provisional application establishes a priority date but is not an examined or granted patent; each must be converted to a non-provisional application within 12 months of its own filing date (by March 25, 2027 and August 11, 2027, respectively) to remain eligible to proceed toward a granted patent, and grant is not guaranteed. Use of this platform grants no patent license, express or implied. Reverse engineering, decompilation, or derivative works are strictly prohibited.
© AariaSec. All rights reserved. All content, interfaces, documentation, and software comprising the AariaSec platform are the exclusive property of AariaSec and its inventors. Unauthorized reproduction, distribution, or modification is strictly prohibited. The AariaSec name, logo, and product names are trademarks of AariaSec.
Authorized use only. AariaSec is licensed solely for authorized security monitoring of AI agents within environments you own or have explicit written permission to monitor. Unauthorized monitoring of third-party systems is prohibited and may violate applicable law, including the California Penal Code § 502 and the federal Computer Fraud and Abuse Act (CFAA).
No legal determinations. Alert findings, risk scores, and panel verdicts are behavioral indicators to assist human security operators. They are not legal determinations of malice, liability, or wrongdoing, and must be reviewed by a qualified human before any enforcement action is taken.
No warranty. The platform is provided "AS IS" without warranty of any kind, express or implied. AariaSec does not warrant that the platform will detect all threats, be error-free, or operate without interruption.
Limitation of liability. To the maximum extent permitted by applicable law, AariaSec's total liability for any claim arising out of or related to the platform shall not exceed the amount you paid for the service in the twelve (12) months preceding the claim. These terms are governed by the laws of the State of California.
Local & hash-only by design. AariaSec runs entirely on your own device. Prompt and response content is reduced to a one-way SHA-256 hash at the moment of interception and the raw bytes are immediately discarded — your agents' prompts and responses are never stored in the clear, never transmitted to AariaSec, and never leave your machine.
What is stored. Only behavioral metadata — timings, counts, destinations, hashes, and risk scores — is retained locally to build each agent's baseline. AariaSec operates no telemetry pipeline and collects no personal data from your deployment.
Optional collective defense. If enabled, deployments may contribute privacy- clean, hash-only threat signatures to a shared reputation feed. These carry no prompt or response content, no personal data, and no organization identity — only anonymized hashes, labels, and counts.