Disclaimer

Disclaimer

Nenya AI Gateway ("Nenya") is provided "as is" for educational and operational use.

1. Best-Effort Security — Not a Guarantee

Nenya includes a regex-based security interceptor designed to detect and block potentially destructive CLI commands (e.g., rm -rf, terraform destroy, DROP TABLE) before they reach your terminal or infrastructure.

This is a best-effort defense layer, not a foolproof guarantee.

Large language models are capable of generating obfuscated, aliased, or otherwise non-obvious command variants that may bypass regex patterns. Examples include but are not limited to:

• Shell aliases and function redefinitions
• Encoded or base64-wrapped commands
• Indirect execution via interpreters (python -c, eval, sh -c)
• Chained commands with non-standard delimiters

No automated filter can fully prevent a determined or hallucinating LLM from producing harmful output. You are solely responsible for validating any command before execution.

2. Autonomous Agent Risk — Zero Liability

Granting autonomous AI agents access to a terminal, cloud infrastructure, or production systems is an inherently dangerous operation with real-world consequences.

Rafael Gumieri, the Nenya contributors, and all affiliated parties are strictly NOT liable for any damage, loss, or harm arising from the use of this software, including but not limited to:

• Data loss or corruption
• Infrastructure demolition or misconfiguration
• Unauthorized cloud billing or resource consumption
• System crashes, denial of service, or availability incidents
• Security breaches caused by commands routed through Nenya
• Any financial, legal, or reputational consequences

Use at your own risk.

3. Human-in-the-Loop Recommendation

We strongly recommend using Nenya in conjunction with human oversight, particularly when agents interact with:

• Production infrastructure — kubectl, helm, terraform, aws, gcloud, az
• Destructive operations — database migrations, certificate rotations, DNS changes
• Billing-sensitive services — auto-scaling groups, spot instances, cross-region replication
• Shared or multi-tenant environments — where blast radius extends beyond your personal workspace

Nenya is designed to reduce risk, not eliminate it. The safest deployment is one where a human operator reviews and approves agent-generated commands before they execute.

4. AI-Assisted Development

This project was rapidly prototyped and built in collaboration with AI engineering tools. Every line of code has been reviewed, tested, and validated by the maintainer. AI tools accelerated the development process; they did not replace human engineering judgment, architectural decisions, or accountability for the final product.

---

Nenya is licensed under the Apache License 2.0. See the LICENSE file for full legal terms.